On 21-01-06 10:05:57, Ben Widawsky wrote:
> On 21-01-06 17:40:14, Jonathan Cameron wrote:
> > On Wed, 6 Jan 2021 13:21:23 +0000
> > Jonathan Cameron <[email protected]> wrote:
> >
> > > On Tue, 5 Jan 2021 08:52:58 -0800
> > > Ben Widawsky <[email protected]> wrote:
> > >
[snip]
>
> I'm sorry you had to debug this. I had fixed this previously and it got lost.
> I'm currently between test applications, so my regression testing isn't great.
>
> I think the fix should be something like this, but I can't easily test at the
> moment:
>
> diff --git a/hw/cxl/cxl-device-utils.c b/hw/cxl/cxl-device-utils.c
> index c515d45d20..b38e9b4c17 100644
> --- a/hw/cxl/cxl-device-utils.c
> +++ b/hw/cxl/cxl-device-utils.c
> @@ -102,6 +102,9 @@ static void mailbox_reg_write(void *opaque, hwaddr
> offset, uint64_t value,
> {
> CXLDeviceState *cxl_dstate = opaque;
>
> + if (offset >= A_CXL_DEV_CMD_PAYLOAD)
> + stn_le_p(cxl_dstate->mbox_reg_state, size, value);
> +
> /*
> * Lock is needed to prevent concurrent writes as well as to prevent
> writes
> * coming in while the firmware is processing. Without background
> commands
>
>
>
+ if (offset >= A_CXL_DEV_CMD_PAYLOAD) {
+ stn_le_p(cxl_dstate->mbox_reg_state, size, value);
+ return;
+ }
+
[snip]
>
