[U-Boot] [PATCH v2 2/3] efi_loader: disk: install FILE_SYSTEM_PROTOCOL only if available

[Heinrich Schuchardt]

On 9/12/19 11:17 AM, AKASHI Takahiro wrote:
> On Thu, Sep 12, 2019 at 10:57:20AM +0200, Heinrich Schuchardt wrote:
>> On 9/12/19 6:51 AM, AKASHI Takahiro wrote:
>>> In the current implementation, EFI_SIMPLEFILE_SYSTEM_PROTOCOL is always
>>> installed to all the partitions even if some of them may house no file
>>> system.
>>>
>>> With this patch, that protocol will be installed only if FAT file system
>>> exists.
>>>
>>> Signed-off-by: AKASHI Takahiro <takahiro.akashi at linaro.org>
>>> ---
>>>  lib/efi_loader/efi_disk.c | 16 +++++++++++++++-
>>>  1 file changed, 15 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/lib/efi_loader/efi_disk.c b/lib/efi_loader/efi_disk.c
>>> index 7a6b06821a47..d72f455901f2 100644
>>> --- a/lib/efi_loader/efi_disk.c
>>> +++ b/lib/efi_loader/efi_disk.c
>>> @@ -9,6 +9,7 @@
>>>  #include <blk.h>
>>>  #include <dm.h>
>>>  #include <efi_loader.h>
>>> +#include <fs.h>
>>>  #include <part.h>
>>>  #include <malloc.h>
>>>
>>> @@ -217,6 +218,19 @@ efi_fs_from_path(struct efi_device_path *full_path)
>>>  	return handler->protocol_interface;
>>>  }
>>>
>>> +static int efi_fs_exists(struct blk_desc *desc, int part)
>>> +{
>>> +	if (fs_set_blk_dev_with_part(desc, part))
>>> +		return 0;
>>> +
>>> +	if (strcmp(fs_get_type_name(), "fat"))
>>
>> Before your patch we could use any supported file system (e.g. EXT2). I
>> see no need for a restriction to FAT. You could compare the string to
>> "unsupported":
>
> No. As far as you want to stick to compliance to UEFI specification,
> "fat" is the only file system supported by UEFI.

In the case of device path node VenHw() there is a direct rule in the
spec indicating how it should be rendered. I have not seen anything in
the UEFI spec saying that you should not support file systems besides
FAT. So there is no compliance issue. I would be reluctant to remove an
existing capability of U-Boot.

>
>> if (!strcmp(fs_get_type_name(), "unsupported"))
>> 	return 0;
>>
>> But wouldn't it be preferable to have a function to access fs_type (in
>> fs/fs.c) directly instead of a string representation?
>
> Agree, but there is no direct function in fs/fs.c.
> I'm reluctant to invent a new function just for this purpose.

In that case we should compare to a string that is defined as constant
in fs.h.

Best regards

Heinrich

>
> -Takahiro Akashi
>
>
>> Otherwise we should convert the string "unsupported" of fstypes[] into a
>> constant in fs.h so that we can be sure we are using the same value.
>>
>> Best regards
>>
>> Heinrich
>>
>>> +		return 0;
>>> +
>>> +	fs_close();
>>> +
>>> +	return 1;
>>> +}
>>> +
>>>  /*
>>>   * Create a handle for a partition or disk
>>>   *
>>> @@ -270,7 +284,7 @@ static efi_status_t efi_disk_add_dev(
>>>  			       diskobj->dp);
>>>  	if (ret != EFI_SUCCESS)
>>>  		return ret;
>>> -	if (part >= 1) {
>>> +	if (part >= 1 && efi_fs_exists(desc, part)) {
>>>  		diskobj->volume = efi_simple_file_system(desc, part,
>>>  							 diskobj->dp);
>>>  		ret = efi_add_protocol(&diskobj->header,
>>>
>>
>