[U-Boot] [PATCH 4/4] efi_loader: unload applications upon Exit()

Heinrich Schuchardt xypron.glpk at gmx.de
Tue May 7 07:07:45 UTC 2019


On 5/7/19 8:37 AM, Takahiro Akashi wrote:
> On Tue, May 07, 2019 at 07:50:48AM +0200, Heinrich Schuchardt wrote:
>> On 5/7/19 6:39 AM, Takahiro Akashi wrote:
>>> On Sat, May 04, 2019 at 10:36:36AM +0200, Heinrich Schuchardt wrote:
>>>> Implement unloading of images in the Exit() boot services:
>>>>
>>>> * unload images that are not yet started,
>>>> * unload started applications,
>>>> * unload drivers returning an error.
>>>>
>>>> Signed-off-by: Heinrich Schuchardt <xypron.glpk at gmx.de>
>>>> ---
>>>>  include/efi_loader.h              |  1 +
>>>>  lib/efi_loader/efi_boottime.c     | 34 ++++++++++++++++++++++++++-----
>>>>  lib/efi_loader/efi_image_loader.c |  2 ++
>>>>  3 files changed, 32 insertions(+), 5 deletions(-)
>>>>
>>>> diff --git a/include/efi_loader.h b/include/efi_loader.h
>>>> index c2a449e5b6..d73c89ac26 100644
>>>> --- a/include/efi_loader.h
>>>> +++ b/include/efi_loader.h
>>>> @@ -237,6 +237,7 @@ struct efi_loaded_image_obj {
>>>>  	struct jmp_buf_data exit_jmp;
>>>>  	EFIAPI efi_status_t (*entry)(efi_handle_t image_handle,
>>>>  				     struct efi_system_table *st);
>>>> +	u16 image_type;
>>>>  };
>>>>
>>>>  /**
>>>> diff --git a/lib/efi_loader/efi_boottime.c b/lib/efi_loader/efi_boottime.c
>>>> index bbcd66caa6..51e0bb2fd5 100644
>>>> --- a/lib/efi_loader/efi_boottime.c
>>>> +++ b/lib/efi_loader/efi_boottime.c
>>>> @@ -13,6 +13,7 @@
>>>>  #include <linux/libfdt_env.h>
>>>>  #include <u-boot/crc.h>
>>>>  #include <bootm.h>
>>>> +#include <pe.h>
>>>>  #include <watchdog.h>
>>>>
>>>>  DECLARE_GLOBAL_DATA_PTR;
>>>> @@ -2798,7 +2799,7 @@ static efi_status_t EFIAPI efi_exit(efi_handle_t image_handle,
>>>>  	 *	 image protocol.
>>>>  	 */
>>>>  	efi_status_t ret;
>>>> -	void *info;
>>>> +	struct efi_loaded_image *loaded_image_protocol;
>>>>  	struct efi_loaded_image_obj *image_obj =
>>>>  		(struct efi_loaded_image_obj *)image_handle;
>>>>
>>>> @@ -2806,13 +2807,33 @@ static efi_status_t EFIAPI efi_exit(efi_handle_t image_handle,
>>>>  		  exit_data_size, exit_data);
>>>>
>>>>  	/* Check parameters */
>>>> -	if (image_handle != current_image)
>>>> +	if (image_handle != current_image) {
>>>
>>> Does this check make sense even for a driver?
>>
>> The check is prescribed in the UEFI specification. See the heading
>> "Status Codes Returned".
>>
>> Multiple binaries may be in status started. If a child image (which may
>> be a driver) calls Exit() with the parent image handle this might cause
>> fancy problems.
>>
>> The lifetime of a driver is LoadImage(), StartImage(), Exit(),
>> UnloadImage(). Typically between StartImage() and Exit() it will install
>> its protocols and between Exit() and UnloadImage() wait for other
>> binaries to call the protocols.
>
> If this is true, that will be fine for a driver.
> But what about 'not started' application? In "Status Codes Returned" of
> Exit(), it says
>   EFI_SUCCESS
>   The image specified by ImageHandle was unloaded. This condition only
>   occurs for images that have been loaded with LoadImage() but have not
>   been started with StartImage().
> In this case, the caller of Exit() is not an application.
>
>>>
>>>> +		ret = EFI_INVALID_PARAMETER;
>>>>  		goto out;
>>>> +	}
>>>>  	ret = EFI_CALL(efi_open_protocol(image_handle, &efi_guid_loaded_image,
>>>> -					 &info, NULL, NULL,
>>>> +					 (void **)&loaded_image_protocol,
>>>> +					 NULL, NULL,
>>>>  					 EFI_OPEN_PROTOCOL_GET_PROTOCOL));
>>>> -	if (ret != EFI_SUCCESS)
>>>> +	if (ret != EFI_SUCCESS) {
>>>
>>> Is this check necessary even when 'header.type' is introduced?
>>
>> I am passing the loaded image protocol to efi_delete_handle().
>
> I don't think that it can be a critical reason.
>
>> In
>> principal a binary might have uninstalled its own loaded image protocol
>> so this call may fail.
>
> If we admit this possibility, there will be bunch of fragile code elsewhere.
>
>>>
>>>> +		ret = EFI_INVALID_PARAMETER;
>>>>  		goto out;
>>>> +	}
>>>> +
>>>> +	/* Unloading of unstarted images */
>>>
>>> 'Unload' sounds odd when we call efi_delete_image(), doesn't it?
>>>
>>>> +	switch (image_obj->header.type) {
>>>> +	case EFI_OBJECT_TYPE_STARTED_IMAGE:
>>>> +		break;
>>>> +	case EFI_OBJECT_TYPE_LOADED_IMAGE:
>>>> +		efi_delete_image(image_obj, loaded_image_protocol);
>>>> +		ret = EFI_SUCCESS;
>>>> +		goto out;
>>>> +	default:
>>>> +		/* Handle does not refer to loaded image */
>>>> +		ret = EFI_INVALID_PARAMETER;
>>>> +		goto out;
>>>> +	}
>>>> +	image_obj->header.type = EFI_OBJECT_TYPE_LOADED_IMAGE;
>>>>
>>>>  	/* Exit data is only foreseen in case of failure. */
>>>>  	if (exit_status != EFI_SUCCESS) {
>>>> @@ -2822,6 +2843,9 @@ static efi_status_t EFIAPI efi_exit(efi_handle_t image_handle,
>>>>  		if (ret != EFI_SUCCESS)
>>>>  			EFI_PRINT("%s: out of memory\n", __func__);
>>>>  	}
>>>> +	if (image_obj->image_type == IMAGE_SUBSYSTEM_EFI_APPLICATION ||
>>>> +	    exit_status != EFI_SUCCESS)
>>>> +		efi_delete_image(image_obj, loaded_image_protocol);
>>>
>>> Why not merge those two efi_delete_image()?
>>
>> I went for readable code. The if statement catching all cases was
>> unreadable to me.
>
> For me, your code is much unreadable.
> Moreover, I remember that you have said, in a review of my patch, that
> we should use "goto" only in error cases.

Good point. So the check must be after handling
EFI_OBJECT_TYPE_LOADED_IMAGE.

I will revise the patch.

Thanks for reviewing.

Best regards

Heinrich

>
> -Takahiro Akashi
>
>>
>> Best regards
>>
>> Heinrich
>>
>>>
>>> -Takahiro Akashi
>>>
>>>
>>>>  	/* Make sure entry/exit counts for EFI world cross-overs match */
>>>>  	EFI_EXIT(exit_status);
>>>> @@ -2837,7 +2861,7 @@ static efi_status_t EFIAPI efi_exit(efi_handle_t image_handle,
>>>>
>>>>  	panic("EFI application exited");
>>>>  out:
>>>> -	return EFI_EXIT(EFI_INVALID_PARAMETER);
>>>> +	return EFI_EXIT(ret);
>>>>  }
>>>>
>>>>  /**
>>>> diff --git a/lib/efi_loader/efi_image_loader.c b/lib/efi_loader/efi_image_loader.c
>>>> index f8092b6202..13541cfa7a 100644
>>>> --- a/lib/efi_loader/efi_image_loader.c
>>>> +++ b/lib/efi_loader/efi_image_loader.c
>>>> @@ -273,6 +273,7 @@ efi_status_t efi_load_pe(struct efi_loaded_image_obj *handle, void *efi,
>>>>  		IMAGE_OPTIONAL_HEADER64 *opt = &nt64->OptionalHeader;
>>>>  		image_base = opt->ImageBase;
>>>>  		efi_set_code_and_data_type(loaded_image_info, opt->Subsystem);
>>>> +		handle->image_type = opt->Subsystem;
>>>>  		efi_reloc = efi_alloc(virt_size,
>>>>  				      loaded_image_info->image_code_type);
>>>>  		if (!efi_reloc) {
>>>> @@ -288,6 +289,7 @@ efi_status_t efi_load_pe(struct efi_loaded_image_obj *handle, void *efi,
>>>>  		IMAGE_OPTIONAL_HEADER32 *opt = &nt->OptionalHeader;
>>>>  		image_base = opt->ImageBase;
>>>>  		efi_set_code_and_data_type(loaded_image_info, opt->Subsystem);
>>>> +		handle->image_type = opt->Subsystem;
>>>>  		efi_reloc = efi_alloc(virt_size,
>>>>  				      loaded_image_info->image_code_type);
>>>>  		if (!efi_reloc) {
>>>> --
>>>> 2.20.1
>>>>
>>>
>>
>



More information about the U-Boot mailing list