Re: [PULL 70/73] hw/pxb-cxl: Support passthrough HDM Decoders unless overridden

[Thomas Huth <thuth@redhat.com>]

On 11/04/2023 12.26, Peter Maydell wrote:
> On Wed, 8 Mar 2023 at 01:14, Michael S. Tsirkin <mst@redhat.com> wrote:
>>
>> From: Jonathan Cameron <Jonathan.Cameron@huawei.com>
>>
>> The CXL r3.0 specification allows for there to be no HDM decoders on CXL
>> Host Bridges if they have only a single root port. Instead, all accesses
>> directed to the host bridge (as specified in CXL Fixed Memory Windows)
>> are assumed to be routed to the single root port.
> 
> Hi; in issue https://gitlab.com/qemu-project/qemu/-/issues/1586
> it's been pointed out that this commit causes an assertion
> failure during 'make check' if you configure with
> --enable-qom-cast-debug. You can repro by doing that and running:
> 
>   qemu-system-i386 -display none -machine q35,cxl=on -device pxb-cxl,bus=pcie.0
> 
> Here's a backtrace:
> 
> Thread 1 "qemu-system-i38" received signal SIGABRT, Aborted.
> __pthread_kill_implementation (no_tid=0, signo=6,
> threadid=140737217810816) at ./nptl/pthread_kill.c:44
> 44      ./nptl/pthread_kill.c: No such file or directory.
> (gdb) bt
> #0  __pthread_kill_implementation (no_tid=0, signo=6,
> threadid=140737217810816) at ./nptl/pthread_kill.c:44
> #1  __pthread_kill_internal (signo=6, threadid=140737217810816) at
> ./nptl/pthread_kill.c:78
> #2  __GI___pthread_kill (threadid=140737217810816,
> signo=signo@entry=6) at ./nptl/pthread_kill.c:89
> #3  0x00007ffff4b1c476 in __GI_raise (sig=sig@entry=6) at
> ../sysdeps/posix/raise.c:26
> #4  0x00007ffff4b027f3 in __GI_abort () at ./stdlib/abort.c:79
> #5  0x0000555555cecfab in object_dynamic_cast_assert
>      (obj=obj@entry=0x555557a70b60, typename=0x555555f80406 "pxb",
> file=0x555555f80357 "../../hw/pci-bridge/pci_expander_bridge.c",
> line=line@entry=55, func=0x555555f8040a "PXB_DEV") at
> ../../qom/object.c:890
> #6  0x00005555559c7bbd in PXB_DEV (obj=0x555557a70b60) at
> ../../hw/pci-bridge/pci_expander_bridge.c:54
> #7  pxb_cxl_dev_reset (dev=0x555557a70b60) at
> ../../hw/pci-bridge/pci_expander_bridge.c:314
> #8  0x00005555559bd624 in pci_qdev_realize (qdev=0x555557a70b60,
> errp=0x7fffffffdd28) at ../../hw/pci/pci.c:2098
> #9  0x0000555555ce8ada in device_set_realized (obj=<optimised out>,
> value=true, errp=0x7fffffffdea8) at ../../hw/core/qdev.c:510
> #10 0x0000555555cf0219 in property_set_bool
>       (obj=obj@entry=0x555557a70b60, v=v@entry=0x555557a727b0,
> name=name@entry=0x55555601db04 "realized", opaque=0x55555687b780,
> errp=errp@entry=0x7fffffffdea8) at ../../qom/object.c:2285
> #11 0x0000555555cee4e5 in object_property_set
>       (obj=obj@entry=0x555557a70b60, name=name@entry=0x55555601db04
> "realized", v=0x555557a727b0, errp=errp@entry=0x7fffffffdea8) at
> ../../qom/object.c:1420
> #12 0x0000555555cf23cd in object_property_set_qobject
>      (obj=obj@entry=0x555557a70b60, name=name@entry=0x55555601db04
> "realized", value=<optimised out>, errp=errp@entry=0x7fffffffdea8) at
> ../../qom/qom-qobject.c:28
> #13 0x0000555555cee93b in object_property_set_bool
> (obj=0x555557a70b60, name=0x55555601db04 "realized", value=<optimised
> out>, errp=0x7fffffffdea8)
>      at ../../qom/object.c:1489
> #14 0x0000555555a6ae42 in qdev_device_add_from_qdict
> (opts=0x555557a6fb40, from_json=false, errp=0x7fffffffdea8,
> errp@entry=0x555556765830 <error_fatal>)
>      at ../../softmmu/qdev-monitor.c:714
> #15 0x0000555555a6b202 in qdev_device_add
> (opts=opts@entry=0x5555568776f0, errp=errp@entry=0x555556765830
> <error_fatal>) at ../../softmmu/qdev-monitor.c:733
> #16 0x0000555555a7367f in device_init_func (opaque=opaque@entry=0x0,
> opts=0x3cd16a, opts@entry=0x5555568776f0, errp=0x6,
> errp@entry=0x555556765830 <error_fatal>)
>      at ../../softmmu/vl.c:1140
> #17 0x0000555555e78331 in qemu_opts_foreach
>      (list=<optimised out>, func=0x555555a73670 <device_init_func>,
> opaque=opaque@entry=0x0, errp=0x555556765830 <error_fatal>) at
> ../../util/qemu-option.c:1135
> #18 0x0000555555a6dd61 in qemu_create_cli_devices () at ../../softmmu/vl.c:2542
> #19 qmp_x_exit_preconfig (errp=<optimised out>) at ../../softmmu/vl.c:2610
> #20 0x0000555555a7177b in qemu_init (argc=<optimised out>,
> argv=<optimised out>) at ../../softmmu/vl.c:3612
> #21 0x000055555587b656 in main (argc=3985770, argv=0x3cd16a) at
> ../../softmmu/main.c:47
> 
> The problem is here:
> 
>> -static void pxb_dev_reset(DeviceState *dev)
>> +static void pxb_cxl_dev_reset(DeviceState *dev)
> 
> This function is called from  pxb_cxl_dev_realize(),
> which is the realize function for TYPE_PXB_CXL_DEVICE.
> That type's parent is TYPE_PCI_DEVICE.
> 
>>   {
>>       CXLHost *cxl = PXB_CXL_DEV(dev)->cxl.cxl_host_bridge;
>>       CXLComponentState *cxl_cstate = &cxl->cxl_cstate;
>> +    PCIHostState *hb = PCI_HOST_BRIDGE(cxl);
>>       uint32_t *reg_state = cxl_cstate->crb.cache_mem_registers;
>>       uint32_t *write_msk = cxl_cstate->crb.cache_mem_regs_write_mask;
>> +    int dsp_count = 0;
>>
>>       cxl_component_register_init_common(reg_state, write_msk, CXL2_ROOT_PORT);
>> -    ARRAY_FIELD_DP32(reg_state, CXL_HDM_DECODER_CAPABILITY, TARGET_COUNT, 8);
>> +    /*
>> +     * The CXL specification allows for host bridges with no HDM decoders
>> +     * if they only have a single root port.
>> +     */
>> +    if (!PXB_DEV(dev)->hdm_for_passthrough) {
> 
> However, here we try to cast the device pointer to PXB_DEV.
> That is not permitted because dev is not of type TYPE_PXB_DEVICE
> (either directly or as a parent class). So if you have the QOM
> debugging enabled then the attempt to cast causes an assertion
> failure.
> 
>> +        dsp_count = pcie_count_ds_ports(hb->bus);
>> +    }
>> +    /* Initial reset will have 0 dsp so wait until > 0 */
>> +    if (dsp_count == 1) {
>> +        cxl->passthrough = true;
>> +        /* Set Capability ID in header to NONE */
>> +        ARRAY_FIELD_DP32(reg_state, CXL_HDM_CAPABILITY_HEADER, ID, 0);
>> +    } else {
>> +        ARRAY_FIELD_DP32(reg_state, CXL_HDM_DECODER_CAPABILITY, TARGET_COUNT,
>> +                         8);
>> +    }
>>   }
> 
> What was the intention here with the type hierarchy?
> Should TYPE_PXB_CXL_DEVICE be a subclass of TYPE_PXB_DEVICE,
> or should the cxl-related functions not be trying to treat
> it as a PXB device ?

Maybe we should simply revert the commit for the time being (once the 8.1 
cycle begins), 'til this has properly been solved, so we can enable 
qom-debug by default again?

  Thomas