[U-Boot] [PATCH v2 1/5] efi_loader: bootmgr: support BootNext and BootCurrent variable behavior

[Heinrich Schuchardt]

On 2/27/19 7:27 AM, AKASHI Takahiro wrote:
> On Wed, Feb 27, 2019 at 07:14:10AM +0100, Heinrich Schuchardt wrote:
>> On 2/27/19 6:47 AM, AKASHI Takahiro wrote:
>>> On Tue, Feb 26, 2019 at 07:57:26PM +0100, Heinrich Schuchardt wrote:
>>>> On 1/15/19 3:54 AM, AKASHI Takahiro wrote:
>>>>> See UEFI v2.7, section 3.1.2 for details of the specification.
>>>>>
>>>>> With my efitool command, you can try as the following:
>>>>>   => efi boot add 1 SHELL ...
>>>>>   => efi boot add 2 HELLO ...
>>>>>   => efi boot order 1 2
>>>>>   => efi bootmgr
>>>>>      (starting SHELL ...)
>>>>>   => efi boot next 2
>>>>>   => efi bootmgr
>>>>>      (starting HELLO ...)
>>>>>   => efi dumpvar
>>>>>   <snip ...>
>>>>>   BootCurrent: {boot,run}(blob)
>>>>>   00000000:  02 00                    ..
>>>>>   BootOrder: {boot,run}(blob)
>>>>>   00000000:  01 00 02 00              ....
>>>>>
>>>>> Signed-off-by: AKASHI Takahiro <takahiro.akashi at linaro.org>
>>>>> ---
>>>>>  lib/efi_loader/efi_bootmgr.c | 34 +++++++++++++++++++++++++++++++++-
>>>>>  1 file changed, 33 insertions(+), 1 deletion(-)
>>>>>
>>>>> diff --git a/lib/efi_loader/efi_bootmgr.c b/lib/efi_loader/efi_bootmgr.c
>>>>> index a095df3f540b..6c5303736dc6 100644
>>>>> --- a/lib/efi_loader/efi_bootmgr.c
>>>>> +++ b/lib/efi_loader/efi_bootmgr.c
>>>>> @@ -145,11 +145,21 @@ static void *try_load_entry(uint16_t n, struct efi_device_path **device_path,
>>>>>  	efi_deserialize_load_option(&lo, load_option);
>>>>>  
>>>>>  	if (lo.attributes & LOAD_OPTION_ACTIVE) {
>>>>> +		u32 attributes;
>>>>>  		efi_status_t ret;
>>>>>  
>>>>>  		debug("%s: trying to load \"%ls\" from %pD\n",
>>>>>  		      __func__, lo.label, lo.file_path);
>>>>>  
>>>>> +		attributes = EFI_VARIABLE_BOOTSERVICE_ACCESS |
>>>>> +			     EFI_VARIABLE_RUNTIME_ACCESS;
>>>>> +		size = sizeof(n);
>>>>> +		ret = rs->set_variable(L"BootCurrent",
>>>>> +				       (efi_guid_t *)&efi_global_variable_guid,
>>>>
>>>> Use EFI_CALL().
>>>
>>> Okay
>>> But as I said somewhere else, it's quite annoying to me that
>>> some efi_xxx requires EFI_CALL(), and others not.
>>> There should have been consistent naming rules.
>>
>> We started with having separate functions like efi_allocate_pages_ext()
>> and efi_allocate_pages(). Then Rob Clark came along and introduced
>> EFI_CALL() in a095aadffa96 and I stopped creating _ext() functions.
>>
>> When running with DEBUG 1 it sometimes is helpful to see which function
>> is calling which other and where errors are originally reported.
>>
>> But I am open to changes in this area.
>>
>>>
>>>> Instead of dereferencing you could directly call
>>>> efi_set_variable().
>>>
>>> Yeah, given that this code is under lib/efi_loader, it may be natural
>>> to use efi_set_variable(). But existing get_var() uses the same style of coding.
>>>
>>> Do you want to change all of the call sites including get_var()?
>>
>> Calling efi_set_variable() directly uses less bytes of code than
>> rs->get_variable() which makes it preferable.
> 
> So is your answer yes, or no?

I would prefer calling efi_get_variable() directly and not to use
rs->get_variable().

> 
>> I have seen that iPXE modifies system->boottime to intercept system
>> calls. The same could be done by an EFI driver to the runtime vectors.
>>
>> In the light of your work on secure boot I think we should not allow an
>> EFI driver to intercept the reading and changing of variables here.
>>
>> We should also rethink it for efidebug.c
> 
> I'm not sure about your concern here, but no doubt efidebug should
> be disabled on production system with secure boot.

Also in efidebug we are creating more runtime code bytes than needed by
using system->runtime->efi_something() or system->boottime->efi_something().

Best regards

Heinrich

> 
> Thanks,
> -Takahiro Akashi
> 
>> Best regards
>>
>> Heinrich
>>
>>>
>>>>> +				       attributes, size, &n);
>>>>> +		if (ret != EFI_SUCCESS)
>>>>> +			goto error;
>>>>> +
>>>>>  		ret = efi_load_image_from_path(lo.file_path, &image);
>>>>>  
>>>>>  		if (ret != EFI_SUCCESS)
>>>>> @@ -173,16 +183,38 @@ error:
>>>>>  void *efi_bootmgr_load(struct efi_device_path **device_path,
>>>>>  		       struct efi_device_path **file_path)
>>>>>  {
>>>>> -	uint16_t *bootorder;
>>>>> +	u16 bootnext, *bootorder;
>>>>>  	efi_uintn_t size;
>>>>>  	void *image = NULL;
>>>>>  	int i, num;
>>>>> +	efi_status_t ret;
>>>>>  
>>>>>  	__efi_entry_check();
>>>>>  
>>>>>  	bs = systab.boottime;
>>>>>  	rs = systab.runtime;
>>>>>  
>>>>> +	/* get BootNext */
>>>>> +	size = sizeof(bootnext);
>>>>> +	ret = rs->get_variable(L"BootNext",
>>>>> +			       (efi_guid_t *)&efi_global_variable_guid,
>>>>> +			       NULL, &size, &bootnext);
>>>>
>>>> You could call efi_get_variable() directly instead of dereferencing rs.
>>>> But anyway you have to use EFI_CALL().
>>>
>>> Ditto
>>>
>>>>> +	if (!bootnext)
>>>>> +		goto run_list;
>>>>
>>>> Goto is acceptable for error handling. But otherwise I would rather
>>>> avoid it.
>>>
>>> Okay with another indentation.
>>>
>>>>> +
>>>>> +	/* delete BootNext */
>>>>> +	ret = rs->set_variable(L"BootNext",
>>>>> +			       (efi_guid_t *)&efi_global_variable_guid,
>>>>> +			       0, 0, &bootnext);
>>>>
>>>> EFI_CALL().
>>>
>>> Thanks,
>>> -Takahiro Akashi
>>>
>>>> Best regards
>>>>
>>>> Heinrich
>>>>
>>>>> +	if (ret != EFI_SUCCESS)
>>>>> +		goto error;
>>>>> +
>>>>> +	image = try_load_entry(bootnext, device_path, file_path);
>>>>> +	if (image)
>>>>> +		goto error;
>>>>> +
>>>>> +run_list:
>>>>> +	/* BootOrder */
>>>>>  	bootorder = get_var(L"BootOrder", &efi_global_variable_guid, &size);
>>>>>  	if (!bootorder)
>>>>>  		goto error;
>>>>>
>>>>
>>>
>>
>