Re: [edk2-devel] [PATCH] ShellPkg/ShellProtocol: Return error code while fail parsing cmd-line

["Philippe Mathieu-Daudé" <philmd@redhat.com>]

Hi Zhichao,

On 12/2/19 12:04 PM, Gao, Zhichao wrote:
>> -----Original Message-----
>> From: Philippe Mathieu-Daudé [mailto:philmd@redhat.com]
>> Sent: Monday, December 2, 2019 6:21 PM
>> To: devel@edk2.groups.io; Gao, Zhichao <zhichao.gao@intel.com>
>> Cc: Ni, Ray <ray.ni@intel.com>; Augustine, Linson <linson.augustine@intel.com>
>> Subject: Re: [edk2-devel] [PATCH] ShellPkg/ShellProtocol: Return error code
>> while fail parsing cmd-line
>>
>> On 12/2/19 1:53 AM, Gao, Zhichao via Groups.Io wrote:
>>> REF: https://bugzilla.tianocore.org/show_bug.cgi?id=2395
>>>
>>> Errors happened in the arguments parsing is not a critical error.
>>> And it would miss the error status code in the release version of shell.
>>> So replace the ASSERT with returning error status code while fail
>>> parsing command-line in UpdateArgcArgv.
>>>
>>> Cc: Ray Ni <ray.ni@intel.com>
>>> Cc: Linson Augustine <linson.augustine@intel.com>
>>> Signed-off-by: Zhichao Gao <zhichao.gao@intel.com>
>>> ---
>>>    ShellPkg/Application/Shell/ShellProtocol.c | 5 ++++-
>>>    1 file changed, 4 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/ShellPkg/Application/Shell/ShellProtocol.c
>>> b/ShellPkg/Application/Shell/ShellProtocol.c
>>> index 5e529b6568..f0362a42d8 100644
>>> --- a/ShellPkg/Application/Shell/ShellProtocol.c
>>> +++ b/ShellPkg/Application/Shell/ShellProtocol.c
>>> @@ -1497,7 +1497,10 @@ InternalShellExecuteDevicePath(
>>>        ShellParamsProtocol.StdOut  =
>> ShellInfoObject.NewShellParametersProtocol->StdOut;
>>>        ShellParamsProtocol.StdErr  =
>> ShellInfoObject.NewShellParametersProtocol->StdErr;
>>>        Status = UpdateArgcArgv(&ShellParamsProtocol, NewCmdLine,
>> Efi_Application, NULL, NULL);
>>> -    ASSERT_EFI_ERROR(Status);
>>> +    if (EFI_ERROR (Status)) {
>>
>> UpdateArgcArgv() is documented to only return EFI_OUT_OF_RESOURCES as
>> error, however it calls ParseCommandLineToArgs() which - also not documented
>> - returns EFI_INVALID_PARAMETER.
>> I suppose this is the "not critical" error this BZ is trying to catch.
>>
>> Should we force Status to EFI_INVALID_PARAMETER before returning, is it safer
>> to return EFI_OUT_OF_RESOURCES if it ever occurs? Should we assert if Status is
>> EFI_OUT_OF_RESOURCES?
> 
> It is fine to return EFI_OUT_OF_RESOURCES of this function as it descripts in its function comments. And all the EFI_OUT_OF_RESOURCES is returned due to the failure of memory allocating.

OK, thanks for clearing that with me.

> And here is an issue of this patch, missing add the return description "EFI_INVALID_PARAMETER" of UpdateArgcArgv  and ParseCommandLineToArgs.
> I would add that in the V2 version.

This is not related to the BZ you are trying to solve, so no need for a 
v2 IMO (and I prepared those patches locally).

Reviewed-by: Philippe Mathieu-Daude <philmd@redhat.com>

>>
>>> +      goto UnloadImage;
>>> +    }
>>> +
>>>        //
>>>        // Replace Argv[0] with the full path of the binary we're executing:
>>>        // If the command line was "foo", the binary might be called "foo.efi".
>>>
>