Re: [Qemu-devel] [PATCH v2 08/18] hw/nvram/fw_cfg: Move fw_cfg_file_slots_allocate() to common_realize()

[Laszlo Ersek <lersek@redhat.com>]

On 03/08/19 02:32, Philippe Mathieu-Daudé wrote:
> Each implementation (I/O and MEM) calls fw_cfg_file_slots_allocate()
> then fw_cfg_common_realize().
> Simplify by moving the fw_cfg_file_slots_allocate() call into
> fw_cfg_common_realize() where it belongs.

The patch does more than that, namely:

> 
> Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
> ---
>  hw/nvram/fw_cfg.c | 17 ++++++++++-------
>  1 file changed, 10 insertions(+), 7 deletions(-)
> 
> diff --git a/hw/nvram/fw_cfg.c b/hw/nvram/fw_cfg.c
> index 0fb020edce..ca58d279a4 100644
> --- a/hw/nvram/fw_cfg.c
> +++ b/hw/nvram/fw_cfg.c
> @@ -929,19 +929,26 @@ static void fw_cfg_machine_ready(struct Notifier *n, void *data)
>      qemu_register_reset(fw_cfg_machine_reset, s);
>  }
>  
> -
> +static void fw_cfg_file_slots_allocate(FWCfgState *s, Error **errp);
>  
>  static void fw_cfg_common_realize(DeviceState *dev, Error **errp)
>  {
>      FWCfgState *s = FW_CFG(dev);
>      MachineState *machine = MACHINE(qdev_get_machine());
>      uint32_t version = FW_CFG_VERSION;
> +    Error *local_err = NULL;
>  
>      if (!fw_cfg_find()) {
>          error_setg(errp, "at most one %s device is permitted", TYPE_FW_CFG);
>          return;
>      }
>  
> +    fw_cfg_file_slots_allocate(s, &local_err);
> +    if (local_err) {
> +        error_propagate(errp, local_err);
> +        return;
> +    }
> +
>      fw_cfg_add_bytes(s, FW_CFG_SIGNATURE, (char *)"QEMU", 4);
>      fw_cfg_add_bytes(s, FW_CFG_UUID, &qemu_uuid, 16);
>      fw_cfg_add_i16(s, FW_CFG_NOGRAPHIC, (uint16_t)!machine->enable_graphics);
> @@ -1108,7 +1115,7 @@ static void fw_cfg_io_realize(DeviceState *dev, Error **errp)
>      FWCfgIoState *s = FW_CFG_IO(dev);
>      Error *local_err = NULL;
>  
> -    fw_cfg_file_slots_allocate(FW_CFG(s), &local_err);
> +    fw_cfg_common_realize(dev, &local_err);
>      if (local_err) {
>          error_propagate(errp, local_err);
>          return;
> @@ -1125,8 +1132,6 @@ static void fw_cfg_io_realize(DeviceState *dev, Error **errp)
>                                &fw_cfg_dma_mem_ops, FW_CFG(s), "fwcfg.dma",
>                                sizeof(dma_addr_t));
>      }
> -
> -    fw_cfg_common_realize(dev, errp);
>  }

it moves the fw_cfg_common_realize() call from the ends of the IO/MEM
realize functions to their middles; in particular to the point before
memory regions and (when appropriate) MMIO resources are initialized.

In other words, the patch reorders all of the current actions in
fw_cfg_common_realize() against said MemoryRegion & SysBusDevice
resource actions.

Why is this safe? (I'm not claiming it is unsafe, but I'd like to see it
explained in the commit message.)

Note: the fw_cfg_common_realize() you are moving around was introduced
in commit 38f3adc34de8 ("fw_cfg: move qdev_init_nofail() from
fw_cfg_init1() to callers", 2017-07-17), by Mark Cave-Ayland. I see Mark
is CC'd already, so I hope he can comment.

Thanks,
Laszlo

>  
>  static void fw_cfg_io_class_init(ObjectClass *klass, void *data)
> @@ -1162,7 +1167,7 @@ static void fw_cfg_mem_realize(DeviceState *dev, Error **errp)
>      const MemoryRegionOps *data_ops = &fw_cfg_data_mem_ops;
>      Error *local_err = NULL;
>  
> -    fw_cfg_file_slots_allocate(FW_CFG(s), &local_err);
> +    fw_cfg_common_realize(dev, &local_err);
>      if (local_err) {
>          error_propagate(errp, local_err);
>          return;
> @@ -1189,8 +1194,6 @@ static void fw_cfg_mem_realize(DeviceState *dev, Error **errp)
>                                sizeof(dma_addr_t));
>          sysbus_init_mmio(sbd, &FW_CFG(s)->dma_iomem);
>      }
> -
> -    fw_cfg_common_realize(dev, errp);
>  }
>  
>  static void fw_cfg_mem_class_init(ObjectClass *klass, void *data)
>