[Qemu-devel] ivshmem assertion failure with EventNotifier

[Qemu-devel] ivshmem assertion failure with EventNotifier

[Cam Macdonell]

Hi Paolo,

I've noticed an assertion error when sending interrupts via ivshmem.
I bisected to this patch.

commit 563027cc0c94aa4846c18f9d665a4c90f8c42ba8
Author: Paolo Bonzini <pbonzini@redhat.com>
Date:   Thu Jul 5 17:16:25 2012 +0200

    ivshmem: use EventNotifier and memory API

    All of ivshmem's usage of eventfd now has a corresponding API in
    EventNotifier.  Simplify the code by using it, and also use the
    memory API consistently to set up and tear down the ioeventfds.

    Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
    Signed-off-by: Avi Kivity <avi@redhat.com>

qemu-system-x86_64: /home/cam/src/git/qemu/memory.c:1244: memory_region_del_even
tfd: Assertion `i != mr->ioeventfd_nb' failed.  This assertion failure
occurs when the eventfd is triggered.

I'll continue to dig around, but can you explain what this assertion
is catching.  Is there an initialization that might be missing?

Thanks,
Cam

Re: [Qemu-devel] ivshmem assertion failure with EventNotifier

[Jan Kiszka]

[-- Attachment #1: Type: text/plain, Size: 1205 bytes --]

On 2012-08-22 06:29, Cam Macdonell wrote:
> Hi Paolo,
> 
> I've noticed an assertion error when sending interrupts via ivshmem.
> I bisected to this patch.
> 
> commit 563027cc0c94aa4846c18f9d665a4c90f8c42ba8
> Author: Paolo Bonzini <pbonzini@redhat.com>
> Date:   Thu Jul 5 17:16:25 2012 +0200
> 
>     ivshmem: use EventNotifier and memory API
> 
>     All of ivshmem's usage of eventfd now has a corresponding API in
>     EventNotifier.  Simplify the code by using it, and also use the
>     memory API consistently to set up and tear down the ioeventfds.
> 
>     Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
>     Signed-off-by: Avi Kivity <avi@redhat.com>
> 
> qemu-system-x86_64: /home/cam/src/git/qemu/memory.c:1244: memory_region_del_even
> tfd: Assertion `i != mr->ioeventfd_nb' failed.  This assertion failure
> occurs when the eventfd is triggered.
> 
> I'll continue to dig around, but can you explain what this assertion
> is catching.  Is there an initialization that might be missing?

Possibly a double-release of the eventfd. The assertion checks if the
parameters provided on del_eventfd match an existing one. Or that
matching is broken.

Jan



[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 262 bytes --]

Re: [Qemu-devel] ivshmem assertion failure with EventNotifier

[Paolo Bonzini]

Il 22/08/2012 06:29, Cam Macdonell ha scritto:
> Hi Paolo,
> 
> I've noticed an assertion error when sending interrupts via ivshmem.
> I bisected to this patch.

Does this help?

diff --git a/hw/ivshmem.c b/hw/ivshmem.c
index b4d65a6..47f2a16 100644
--- a/hw/ivshmem.c
+++ b/hw/ivshmem.c
@@ -366,6 +366,10 @@ static void close_guest_eventfds(IVShmemState *s,
int posn)
 {
     int i, guest_curr_max;

+    if (!ivshmem_has_feature(s, IVSHMEM_IOEVENTFD)) {
+        return;
+    }
+
     guest_curr_max = s->peers[posn].nb_eventfds;

     memory_region_transaction_begin();

Paolo

Re: [Qemu-devel] ivshmem assertion failure with EventNotifier

[Cam Macdonell]

On Wed, Aug 22, 2012 at 6:01 AM, Paolo Bonzini <pbonzini@redhat.com> wrote:
> Il 22/08/2012 06:29, Cam Macdonell ha scritto:
>> Hi Paolo,
>>
>> I've noticed an assertion error when sending interrupts via ivshmem.
>> I bisected to this patch.
>
> Does this help?

Yes, that solves it.

>
> diff --git a/hw/ivshmem.c b/hw/ivshmem.c
> index b4d65a6..47f2a16 100644
> --- a/hw/ivshmem.c
> +++ b/hw/ivshmem.c
> @@ -366,6 +366,10 @@ static void close_guest_eventfds(IVShmemState *s,
> int posn)
>  {
>      int i, guest_curr_max;
>
> +    if (!ivshmem_has_feature(s, IVSHMEM_IOEVENTFD)) {
> +        return;
> +    }
> +
>      guest_curr_max = s->peers[posn].nb_eventfds;
>
>      memory_region_transaction_begin();
>
> Paolo
>