Re: [PATCH v7 1/5] hw/nvram/fw_cfg: Add the FW_CFG_DATA_GENERATOR interface

[Philippe Mathieu-Daudé]

On 5/29/20 11:09 AM, Laszlo Ersek wrote:
> On 05/28/20 19:31, Philippe Mathieu-Daudé wrote:
>> The FW_CFG_DATA_GENERATOR allows any object to produce
>> blob of data consumable by the fw_cfg device.
>>
>> Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
>> ---
>> v7: addressed Laszlo's comments
>> - fixed typos in description
>> - return size_t instead of ssize_t; 0 for error
>> - do not use 1-letter variable names
>> - do not open-code 'fw_cfg-data-generator'
>> - cast g_memdup() size argument as 'guint'
>> - improved documentation
>> ---
>>  docs/specs/fw_cfg.txt     |  9 ++++++-
>>  include/hw/nvram/fw_cfg.h | 52 +++++++++++++++++++++++++++++++++++++++
>>  hw/nvram/fw_cfg.c         | 31 +++++++++++++++++++++++
>>  3 files changed, 91 insertions(+), 1 deletion(-)
>>
>> diff --git a/docs/specs/fw_cfg.txt b/docs/specs/fw_cfg.txt
>> index 8f1ebc66fa..bc16daa38a 100644
>> --- a/docs/specs/fw_cfg.txt
>> +++ b/docs/specs/fw_cfg.txt
>> @@ -219,7 +219,7 @@ To check the result, read the "control" field:
>>  
>>  = Externally Provided Items =
>>  
>> -As of v2.4, "file" fw_cfg items (i.e., items with selector keys above
>> +Since v2.4, "file" fw_cfg items (i.e., items with selector keys above
>>  FW_CFG_FILE_FIRST, and with a corresponding entry in the fw_cfg file
>>  directory structure) may be inserted via the QEMU command line, using
>>  the following syntax:
>> @@ -230,6 +230,13 @@ Or
>>  
>>      -fw_cfg [name=]<item_name>,string=<string>
>>  
>> +Since v5.1, QEMU allows some objects to generate fw_cfg-specific content,
>> +the content is then associated with a "file" item using the 'gen_id' option
>> +in the command line, using the following syntax:
>> +
>> +    -object <generator-type>,id=<generated_id>,[generator-specific-options] 
>> \
>> +    -fw_cfg [name=]<item_name>,gen_id=<generated_id>
>> +
>>  See QEMU man page for more documentation.
>>  
>>  Using item_name with plain ASCII characters only is recommended.
> 
> I've looked at this hunk with a larger context, and I think it's really
> good.
> 
>> diff --git a/include/hw/nvram/fw_cfg.h b/include/hw/nvram/fw_cfg.h
>> index 25d9307018..8fbf2446c1 100644
>> --- a/include/hw/nvram/fw_cfg.h
>> +++ b/include/hw/nvram/fw_cfg.h
>> @@ -9,11 +9,43 @@
>>  #define TYPE_FW_CFG     "fw_cfg"
>>  #define TYPE_FW_CFG_IO  "fw_cfg_io"
>>  #define TYPE_FW_CFG_MEM "fw_cfg_mem"
>> +#define TYPE_FW_CFG_DATA_GENERATOR_INTERFACE "fw_cfg-data-generator"
>>  
>>  #define FW_CFG(obj)     OBJECT_CHECK(FWCfgState,    (obj), TYPE_FW_CFG)
>>  #define FW_CFG_IO(obj)  OBJECT_CHECK(FWCfgIoState,  (obj), TYPE_FW_CFG_IO)
>>  #define FW_CFG_MEM(obj) OBJECT_CHECK(FWCfgMemState, (obj), TYPE_FW_CFG_MEM)
>>  
>> +#define FW_CFG_DATA_GENERATOR_CLASS(class) \
>> +    OBJECT_CLASS_CHECK(FWCfgDataGeneratorClass, (class), \
>> +                       TYPE_FW_CFG_DATA_GENERATOR_INTERFACE)
>> +#define FW_CFG_DATA_GENERATOR_GET_CLASS(obj) \
>> +    OBJECT_GET_CLASS(FWCfgDataGeneratorClass, (obj), \
>> +                     TYPE_FW_CFG_DATA_GENERATOR_INTERFACE)
>> +
>> +typedef struct FWCfgDataGeneratorClass {
>> +    /*< private >*/
>> +    InterfaceClass parent_class;
>> +    /*< public >*/
>> +
>> +    /**
>> +     * get_data:
>> +     * @obj: the object implementing this interface
>> +     *
>> +     * Returns: pointer to start of the generated item data
>> +     *
>> +     * The returned pointer is a QObject weak reference, @obj owns
>> +     * the reference and may free it at any time in the future.
>> +     */
>> +    const void *(*get_data)(Object *obj);
>> +    /**
>> +     * get_length:
>> +     * @obj: the object implementing this interface
>> +     *
>> +     * Returns: the size of the generated item data in bytes
>> +     */
>> +    size_t (*get_length)(Object *obj);
>> +} FWCfgDataGeneratorClass;
>> +
>>  typedef struct fw_cfg_file FWCfgFile;
>>  
>>  #define FW_CFG_ORDER_OVERRIDE_VGA    70
>> @@ -263,6 +295,26 @@ void fw_cfg_add_file_callback(FWCfgState *s, const char 
>> *filename,
>>  void *fw_cfg_modify_file(FWCfgState *s, const char *filename, void *data,
>>                           size_t len);
>>  
>> +/**
>> + * fw_cfg_add_from_generator:
>> + * @s: fw_cfg device being modified
>> + * @filename: name of new fw_cfg file item
>> + * @gen_id: name of object implementing FW_CFG_DATA_GENERATOR interface
>> + * @errp: pointer to a NULL initialized error object
>> + *
>> + * Add a new NAMED fw_cfg item with the content generated from the
>> + * @gen_id object. The data generated by the @gen_id object/ is copied
> 
> (1) typo: "object/" (possibly a copy-paste error from my v6 review)
> 
>> + * into the data structure of the fw_cfg device.
>> + * The next available (unused) selector key starting at FW_CFG_FILE_FIRST
>> + * will be used; also, a new entry will be added to the file directory
>> + * structure residing at key value FW_CFG_FILE_DIR, containing the item 
>> name,
>> + * data size, and assigned selector key value.
>> + *
>> + * Returns: the size of the device tree image on success, or 0 on errors.
> 
> (2) typo (probably another copy-paste error):
> 
> s/device tree image/generated item data/
> 
>> + */
>> +size_t fw_cfg_add_from_generator(FWCfgState *s, const char *filename,
>> +                                 const char *gen_id, Error **errp);
>> +
>>  FWCfgState *fw_cfg_init_io_dma(uint32_t iobase, uint32_t dma_iobase,
>>                                  AddressSpace *dma_as);
>>  FWCfgState *fw_cfg_init_io(uint32_t iobase);
>> diff --git a/hw/nvram/fw_cfg.c b/hw/nvram/fw_cfg.c
>> index 8dd50c2c72..6d2fa13042 100644
>> --- a/hw/nvram/fw_cfg.c
>> +++ b/hw/nvram/fw_cfg.c
>> @@ -1032,6 +1032,31 @@ void *fw_cfg_modify_file(FWCfgState *s, const char 
>> *filename,
>>      return NULL;
>>  }
>>  
>> +size_t fw_cfg_add_from_generator(FWCfgState *s, const char *filename,
>> +                                 const char *gen_id, Error **errp)
>> +{
>> +    FWCfgDataGeneratorClass *klass;
>> +    Object *obj;
>> +    size_t size;
>> +
>> +    obj = object_resolve_path_component(object_get_objects_root(), gen_id);
>> +    if (!obj) {
>> +        error_setg(errp, "Cannot find object ID %s", gen_id);
>> +        return 0;
>> +    }
>> +    if (!object_dynamic_cast(obj, TYPE_FW_CFG_DATA_GENERATOR_INTERFACE)) {
>> +        error_setg(errp, "Object '%s' is not a '%s' subclass",
>> +                   TYPE_FW_CFG_DATA_GENERATOR_INTERFACE, gen_id);
> 
> (3) the order of the last two arguments is wrong; please reverse it.

Oops...

> 
>> +        return 0;
>> +    }
>> +    klass = FW_CFG_DATA_GENERATOR_GET_CLASS(obj);
>> +    size = klass->get_length(obj);
> 
> (4) I think we should not call fw_cfg_add_file() if get_length() returns
> zero.
> 
> AFAICS, fw_cfg_add_file() will happily create a useless zero-size fw_cfg
> entry, but then the same zero size that we return from
> fw_cfg_add_from_generator() will indicate failure to the caller (and we
> will also not have set "errp").

Good catch.

> 
> I think we need something like:
> 
>     if (size == 0) {
>         error_setg(errp, "Object '%s' failed to generate fw_cfg data",
>                    gen_id);
>         return 0;
>     }
> 
> That will also prevent an awkward call to g_memdup() with size==0.
> 
> 
> Now, looking at patches #4 and #5, I think if no suitable cipher suite
> is found, then qcrypto_tls_cipher_suites_get_data() will return NULL,
> and qcrypto_tls_cipher_suites_get_length() will return zero. In turn, my
> suggestion (4) here will prevent QEMU from starting up.
> 
> And I think that is just fine. The corresponding firmware function is
> SetCipherSuites() in
> "OvmfPkg/Library/TlsAuthConfigLib/TlsAuthConfigLib.c", and if it sees a
> zero-sized "etc/edk2/https/ciphers" fw_cfg file, it will hang the
> firmware, *on purpose*. (See the reason explained in comments there.)
> 
> Thus, it's much better if we prevent QEMU from starting up in the first
> place.
> 
>> +    fw_cfg_add_file(s, filename, g_memdup(klass->get_data(obj), 
>> (guint)size),
>> +                    size);
>> +
>> +    return size;
>> +}
>> +
>>  static void fw_cfg_machine_reset(void *opaque)
>>  {
>>      MachineClass *mc = MACHINE_GET_CLASS(qdev_get_machine());
>> @@ -1333,12 +1358,18 @@ static const TypeInfo fw_cfg_mem_info = {
>>      .class_init    = fw_cfg_mem_class_init,
>>  };
>>  
>> +static const TypeInfo fw_cfg_data_generator_interface_info = {
>> +    .name = TYPE_FW_CFG_DATA_GENERATOR_INTERFACE,
>> +    .parent = TYPE_INTERFACE,
>> +    .class_size = sizeof(FWCfgDataGeneratorClass),
>> +};
>>  
>>  static void fw_cfg_register_types(void)
>>  {
>>      type_register_static(&fw_cfg_info);
>>      type_register_static(&fw_cfg_io_info);
>>      type_register_static(&fw_cfg_mem_info);
>> +    type_register_static(&fw_cfg_data_generator_interface_info);
>>  }
>>  
>>  type_init(fw_cfg_register_types)
>>
> 
> With (1) through (4) fixed:
> 
> Reviewed-by: Laszlo Ersek <lersek@redhat.com>

Will fix, thanks!

> 
> Thanks
> Laszlo
>