Re: [RFC PATCH] tests/qtest: properly initialise the vring used idx

["Alex Bennée" <alex.bennee@linaro.org>]

Peter Maydell <peter.maydell@linaro.org> writes:

> On Wed, 6 Apr 2022 at 21:07, Alex Bennée <alex.bennee@linaro.org> wrote:
>>
>>
>> Peter Maydell <peter.maydell@linaro.org> writes:
>> > Guest memory is generally zero at startup. Do we manage to
>> > hit the bit of memory at the start of the virt machine's RAM
>> > where we store the DTB ? (As you say, initializing the data
>> > structures is the right thing anyway.)
>>
>> I don't know - where is the DTB loaded?
>
> Start of RAM (that's physaddr 0x4000_0000). The thing I'm not sure
> about is whether these qtests go through the code in hw/arm/boot.c
> that loads the DTB into guest RAM or not.

Yes because it's linked to the machine creation:

Thread 1 hit Breakpoint 1, arm_load_dtb (addr=1073741824, binfo=binfo@entry=0x55bc4ce26970, addr_limit=0, as=as@entry=0x55bc4d119c50, ms=ms@entry=0x55bc4ce26800) at ../../hw/arm/boot.c:534
534     {
(rr) bt
#0  arm_load_dtb (addr=1073741824, binfo=binfo@entry=0x55bc4ce26970, addr_limit=0, as=as@entry=0x55bc4d119c50, ms=ms@entry=0x55bc4ce26800) at ../../hw/arm/boot.c:534
#1  0x000055bc4a9f7ded in virt_machine_done (notifier=0x55bc4ce26910, data=<optimized out>) at ../../hw/arm/virt.c:1637
#2  0x000055bc4aebc807 in notifier_list_notify (list=list@entry=0x55bc4b5f8b20 <machine_init_done_notifiers>, data=data@entry=0x0) at ../../util/notify.c:39
#3  0x000055bc4a7f82db in qdev_machine_creation_done () at ../../hw/core/machine.c:1235
#4  0x000055bc4a744b19 in qemu_machine_creation_done () at ../../softmmu/vl.c:2725
#5  qmp_x_exit_preconfig (errp=<optimized out>) at ../../softmmu/vl.c:2748
#6  0x000055bc4a748a14 in qmp_x_exit_preconfig (errp=<optimized out>) at ../../softmmu/vl.c:2741
#7  qemu_init (argc=<optimized out>, argv=<optimized out>, envp=<optimized out>) at ../../softmmu/vl.c:3776
#8  0x000055bc4a6de639 in main (argc=<optimized out>, argv=<optimized out>, envp=<optimized out>) at ../../softmmu/main.c:49

(ION: yay, I can capture qtest runs in rr now ;-)

>
>> Currently we are using the first
>> couple of pages in qtest because that where the qtest allocater is
>> initialised:
>>
>>   static void *qos_create_machine_arm_virt(QTestState *qts)
>>   {
>>       QVirtMachine *machine = g_new0(QVirtMachine, 1);
>>
>>       alloc_init(&machine->alloc, 0,
>>                  ARM_VIRT_RAM_ADDR,
>>                  ARM_VIRT_RAM_ADDR + ARM_VIRT_RAM_SIZE,
>>                  ARM_PAGE_SIZE);
>>       qvirtio_mmio_init_device(&machine->virtio_mmio, qts, VIRTIO_MMIO_BASE_ADDR,
>>                                VIRTIO_MMIO_SIZE);
>>
>>       qos_create_generic_pcihost(&machine->bridge, qts, &machine->alloc);
>>
>>       machine->obj.get_device = virt_get_device;
>>       machine->obj.get_driver = virt_get_driver;
>>       machine->obj.destructor = virt_destructor;
>>       return machine;
>>   }
>>
>> I don't know if there is a more sane piece of memory we should be using.
>
> The first part of RAM is fine, it's just you can't assume it's
> all zeroes :-)
>
> -- PMM


-- 
Alex Bennée