Re: [PATCH] vmbus: remove hv_event_tasklet_disable/enable

From: Stephen Hemminger
Date: Fri Mar 03 2017 - 12:12:23 EST


On Fri, 03 Mar 2017 17:40:47 +0100
Vitaly Kuznetsov <vkuznets@xxxxxxxxxx> wrote:

> Dexuan Cui <decui@xxxxxxxxxxxxx> writes:
>
> > With the recent introduction of per-channel tasklet, we need to update
> > the way we handle the 3 concurrency issues:
> >
> > 1. hv_process_channel_removal -> percpu_channel_deq vs.
> > vmbus_chan_sched -> list_for_each_entry(..., percpu_list);
> >
> > 2. vmbus_process_offer -> percpu_channel_enq/deq vs. vmbus_chan_sched.
> >
> > 3. vmbus_close_internal vs. the per-channel tasklet vmbus_on_event;
> >
> > The first 2 issues can be handled by Stephen's recent patch
> > "vmbus: use rcu for per-cpu channel list", and the third issue
> > can be handled by calling tasklet_disable in vmbus_close_internal here.
> >
> > We don't need the original hv_event_tasklet_disable/enable since we
> > now use per-channel tasklet instead of the previous per-CPU tasklet,
> > and actually we must remove them due to the side effect now:
> > vmbus_process_offer -> hv_event_tasklet_enable -> tasklet_schedule will
> > start the per-channel callback prematurely, cauing NULL dereferencing
> > (the channel may haven't been properly configured to run the callback yet).
> >
> > Fixes: 631e63a9f346 ("vmbus: change to per channel tasklet")
> >
> > Signed-off-by: Dexuan Cui <decui@xxxxxxxxxxxxx>
> > Cc: "K. Y. Srinivasan" <kys@xxxxxxxxxxxxx>
> > Cc: Haiyang Zhang <haiyangz@xxxxxxxxxxxxx>
> > Cc: Stephen Hemminger <sthemmin@xxxxxxxxxxxxx>
>
> Tested-by: Vitaly Kuznetsov <vkuznets@xxxxxxxxxx>
>
> This patch fixes the following crash on boot:
>
> [ 1.451648] BUG: unable to handle kernel NULL pointer dereference at 0000000000000004
> [ 1.452255] IP: netvsc_channel_cb+0x90/0x7b0 [hv_netvsc]
> [ 1.452255] PGD 0
> [ 1.452255]
> [ 1.452255] Oops: 0000 [#1] SMP
> [ 1.452255] Modules linked in: hv_storvsc hv_netvsc(+) scsi_transport_fc hyperv_fb hyperv_keyboard hid_hyperv hv_vmbus
> [ 1.452255] CPU: 1 PID: 19 Comm: ksoftirqd/1 Not tainted 4.10.0_test+ #911
> [ 1.452255] Hardware name: Microsoft Corporation Virtual Machine/Virtual Machine, BIOS Hyper-V UEFI Release v1.0 11/26/2012
> [ 1.452255] task: ffff880007fd2b00 task.stack: ffffc90000e34000
> [ 1.452255] RIP: 0010:netvsc_channel_cb+0x90/0x7b0 [hv_netvsc]
> ...
> [ 1.452255] Call Trace:
> [ 1.452255] vmbus_on_event+0x22/0x90 [hv_vmbus]
> [ 1.452255] tasklet_action+0x5e/0x110
> [ 1.452255] __do_softirq+0x104/0x2af
> [ 1.452255] run_ksoftirqd+0x29/0x40
> ...
> [ 1.548068] RIP: netvsc_channel_cb+0x90/0x7b0 [hv_netvsc] RSP: ffffc90000e37d88
> [ 1.548068] CR2: 0000000000000004
> [ 1.548068] ---[ end trace 601fd9d6588b21e5 ]---
> [ 1.548068] Kernel panic - not syncing: Fatal exception in interrupt
> [ 1.548068] Kernel Offset: disabled
> [ 1.548068] ---[ end Kernel panic - not syncing: Fatal exception in interrupt
> [ 1.572155] ------------[ cut here ]------------
>
> The crash is not imminent but it happens pretty often on boot, I think
> we need to push it to 4.11.
>

Agree that this needs to be in 4.11, but when NAPI is merged it also will not happen.