From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail-wm0-x229.google.com (mail-wm0-x229.google.com [IPv6:2a00:1450:400c:c09::229]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ml01.01.org (Postfix) with ESMTPS id F3AFF21D1B2AB for ; Wed, 19 Jul 2017 14:56:07 -0700 (PDT) Received: by mail-wm0-x229.google.com with SMTP id g127so10839857wmd.0 for ; Wed, 19 Jul 2017 14:58:03 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=linaro.org; s=google; h=date:from:to:cc:subject:message-id:references:mime-version :content-disposition:in-reply-to:user-agent; bh=nZjRQkF5sx2eo0k3otpQ5uooSE7H8Idk0ZxErT5EGNs=; b=IKldeYN3M+MTIGjlFV7IZsAEeO1CjJCeZ+zDKnvMfItBkzsLSQY8Cmp3WCs2+CGWc4 57B8lhfJeoEmbrF0mgphJAtz4P2kdaDpGzf8cDReefTTrKK8gA+M0qu2SuMATMXVsA+U gQT4cj3zSszx+DLhqAZDtyWSYaDSDSS7iCCes= X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:date:from:to:cc:subject:message-id:references :mime-version:content-disposition:in-reply-to:user-agent; bh=nZjRQkF5sx2eo0k3otpQ5uooSE7H8Idk0ZxErT5EGNs=; b=ZCWrBH5FwVEns5JZb/oDWWCCdkdxawUO3pHWQcZm/pLcRDoGWsYST2P+KXghe7UefC YXuB5iA7pKHo2yCx8f2OiaJ2cEyerMTJPHIcrB+3GAkNxPLON7bDi9rTCOx7qdev+zIb JKKWRca4yXXJ3BKkxDf81fYhpAjQYRxPs4otGOQ3CHiCk5kZjsQbpBUz0A60WDN3rM1Z 6PYhZw3o5QW1P+G01cR5VxxyeINBGAgtF3ShFohCs9Vsa+Y0Q885ZJwLKW4mPZvDWy6L or8nGovg0K6ZmD6fqxw9d0X2qCrfqPrqv+r34dgArVuW1UOA35evLFCFqkIid0ipwO9t MGhA== X-Gm-Message-State: AIVw111Dres4paKG1lQxxnORxBavRMW9Tmy7c4G4nBYKRIZT9BA2PEGQ /ifItU84mB/IfkMw X-Received: by 10.28.11.134 with SMTP id 128mr933774wml.79.1500501481216; Wed, 19 Jul 2017 14:58:01 -0700 (PDT) Received: from bivouac.eciton.net (bivouac.eciton.net. [2a00:1098:0:86:1000:23:0:2]) by smtp.gmail.com with ESMTPSA id z190sm868480wmz.31.2017.07.19.14.58.00 (version=TLS1_2 cipher=ECDHE-RSA-AES128-GCM-SHA256 bits=128/128); Wed, 19 Jul 2017 14:58:00 -0700 (PDT) Date: Wed, 19 Jul 2017 22:57:58 +0100 From: Leif Lindholm To: Jun Nie Cc: Haojian Zhuang , Ard Biesheuvel , edk2-devel@lists.01.org, linaro-uefi@lists.linaro.org, Shawn Guo , Jason Liu Message-ID: <20170719215758.GA1501@bivouac.eciton.net> References: <1499347746-16786-1-git-send-email-jun.nie@linaro.org> <20170718160420.GR26676@bivouac.eciton.net> MIME-Version: 1.0 In-Reply-To: User-Agent: Mutt/1.5.23 (2014-03-12) Subject: Re: [PATCH v2 1/2] EmbeddedPkg/AndroidBoot: boot android kernel from storage X-BeenThere: edk2-devel@lists.01.org X-Mailman-Version: 2.1.22 Precedence: list List-Id: EDK II Development List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Wed, 19 Jul 2017 21:56:08 -0000 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline On Wed, Jul 19, 2017 at 06:00:04PM +0800, Jun Nie wrote: > 2017-07-19 0:04 GMT+08:00 Leif Lindholm : > > On Thu, Jul 06, 2017 at 09:29:05PM +0800, Jun Nie wrote: > >> Add an android kernel loader that could load kernel from storage > >> device. > > > > UEFI can already load a kernel (with the EFI stub) from a storage > > device. Please explain in the commit message how this support differs > > from that. > > OK, will add description for addtitional cmdline/dtb/ramfs support besides > kernel that is introduced by Android boot header. Allright, after having read through your responses, I think I understand what is going on. But yeah, a link to some documentation of this Android boot image format would be useful. > > What relation does this code have to AndroidFastbootApp? > > It does not relate to AndroidFastbootApp directly because fastboot is tools > for image download/flash/etc. But both apps comply to Google's spec > and share some common data definition. Got it, thanks. > >> This patch is from Haojian's code. > > > > Could you put a link to the origin of the code (preferably the > > specific commit)? > Will do. > > > >> The minor change > >> is that alternative dtb is searched in second loader binary of > >> Android bootimage if dtb is not found after Linux kernel. > >> > >> Contributed-under: TianoCore Contribution Agreement 1.0 > >> Signed-off-by: Jun Nie > >> --- > >> .../Application/AndroidBoot/AndroidBootApp.c | 129 ++++++++ > >> .../Application/AndroidBoot/AndroidBootApp.inf | 64 ++++ > >> EmbeddedPkg/Include/Library/AbootimgLib.h | 65 ++++ > >> EmbeddedPkg/Include/Protocol/Abootimg.h | 47 +++ > >> EmbeddedPkg/Library/AbootimgLib/AbootimgLib.c | 350 +++++++++++++++++++++ > >> EmbeddedPkg/Library/AbootimgLib/AbootimgLib.inf | 48 +++ > > > > For proper CamelCase, I think the library name should be > > AndroidBootImageLib, and the filenames (and macros) updated similarly. > > Will do. > > > >> 6 files changed, 703 insertions(+) > >> create mode 100644 EmbeddedPkg/Application/AndroidBoot/AndroidBootApp.c > >> create mode 100644 EmbeddedPkg/Application/AndroidBoot/AndroidBootApp.inf > >> create mode 100644 EmbeddedPkg/Include/Library/AbootimgLib.h > >> create mode 100644 EmbeddedPkg/Include/Protocol/Abootimg.h > >> create mode 100644 EmbeddedPkg/Library/AbootimgLib/AbootimgLib.c > >> create mode 100644 EmbeddedPkg/Library/AbootimgLib/AbootimgLib.inf > >> > >> diff --git a/EmbeddedPkg/Application/AndroidBoot/AndroidBootApp.c b/EmbeddedPkg/Application/AndroidBoot/AndroidBootApp.c > >> new file mode 100644 > >> index 0000000..9ed931b > >> --- /dev/null > >> +++ b/EmbeddedPkg/Application/AndroidBoot/AndroidBootApp.c > >> @@ -0,0 +1,129 @@ > >> +/** @file > >> + > >> + Copyright (c) 2013-2014, ARM Ltd. All rights reserved.
> >> + Copyright (c) 2017, Linaro. All rights reserved. > >> + > >> + This program and the accompanying materials > >> + are licensed and made available under the terms and conditions of the BSD License > >> + which accompanies this distribution. The full text of the license may be found at > >> + http://opensource.org/licenses/bsd-license.php > >> + > >> + THE PROGRAM IS DISTRIBUTED UNDER THE BSD LICENSE ON AN "AS IS" BASIS, > >> + WITHOUT WARRANTIES OR REPRESENTATIONS OF ANY KIND, EITHER EXPRESS OR IMPLIED. > >> + > >> +**/ > >> + > >> +#include > >> +#include > >> +#include > >> +#include > >> +#include > >> +#include > >> +#include > >> + > >> +#include > >> +#include > >> + > >> +#define IS_DEVICE_PATH_NODE(node,type,subtype) (((node)->Type == (type)) && ((node)->SubType == (subtype))) > > > > Would prefer for this to be moved into a common header file (BdsLib.h > > ?) rather than copied into multiple .c files. > > Will do. > > > >> + > >> +#define ALIGN(x, a) (((x) + ((a) - 1)) & ~((a) - 1)) > > > > Rather use one of the ALIGN_ macros from Base.h. > > > > Will do. > >> + > >> +EFI_STATUS > >> +EFIAPI > >> +AndroidBootAppEntryPoint ( > >> + IN EFI_HANDLE ImageHandle, > >> + IN EFI_SYSTEM_TABLE *SystemTable > >> + ) > >> +{ > >> + EFI_STATUS Status; > >> + CHAR16 *BootPathStr; > >> + EFI_DEVICE_PATH_FROM_TEXT_PROTOCOL *EfiDevicePathFromTextProtocol; > >> + EFI_DEVICE_PATH *DevicePath; > >> + EFI_DEVICE_PATH_PROTOCOL *Node, *NextNode; > >> + EFI_BLOCK_IO_PROTOCOL *BlockIo; > >> + UINT32 MediaId, BlockSize; > >> + VOID *Buffer; > >> + EFI_HANDLE Handle; > >> + UINTN Size; > >> + > >> + BootPathStr = (CHAR16 *)PcdGetPtr (PcdAndroidBootDevicePath); > >> + ASSERT (BootPathStr != NULL); > >> + Status = gBS->LocateProtocol (&gEfiDevicePathFromTextProtocolGuid, NULL, (VOID **)&EfiDevicePathFromTextProtocol); > >> + ASSERT_EFI_ERROR(Status); > >> + DevicePath = (EFI_DEVICE_PATH *)EfiDevicePathFromTextProtocol->ConvertTextToDevicePath (BootPathStr); > >> + ASSERT (DevicePath != NULL); > >> + > >> + /* Find DevicePath node of Partition */ > >> + NextNode = DevicePath; > >> + while (1) { > >> + Node = NextNode; > >> + if (IS_DEVICE_PATH_NODE (Node, MEDIA_DEVICE_PATH, MEDIA_HARDDRIVE_DP)) { > >> + break; > >> + } > >> + NextNode = NextDevicePathNode (Node); > >> + } > >> + > >> + Status = gBS->LocateDevicePath (&gEfiDevicePathProtocolGuid, &DevicePath, &Handle); > >> + if (EFI_ERROR (Status)) { > >> + return Status; > >> + } > >> + > >> + Status = gBS->OpenProtocol ( > >> + Handle, > >> + &gEfiBlockIoProtocolGuid, > >> + (VOID **) &BlockIo, > >> + gImageHandle, > >> + NULL, > >> + EFI_OPEN_PROTOCOL_GET_PROTOCOL > >> + ); > >> + if (EFI_ERROR (Status)) { > >> + DEBUG ((EFI_D_ERROR, "Failed to get BlockIo: %r\n", Status)); > >> + return Status; > >> + } > >> + > >> + MediaId = BlockIo->Media->MediaId; > >> + BlockSize = BlockIo->Media->BlockSize; > >> + Buffer = AllocatePages (1); > > > > I dislike magic numbers even where the code is reasonably clear. > > Could this be EFI_SIZE_TO_PAGES (ANDROID_BOOTIMG_HEADER) instead of > > "1"? > > Will do. > > > >> + if (Buffer == NULL) { > >> + return EFI_BUFFER_TOO_SMALL; > >> + } > >> + /* Load header of boot.img */ > >> + Status = BlockIo->ReadBlocks ( > >> + BlockIo, > >> + MediaId, > >> + 0, > >> + BlockSize, > >> + Buffer > >> + ); > >> + Status = AbootimgGetImgSize (Buffer, &Size); > >> + if (EFI_ERROR (Status)) { > >> + DEBUG ((EFI_D_ERROR, "Failed to get Abootimg Size: %r\n", Status)); > >> + return Status; > >> + } > >> + Size = ALIGN (Size, BlockSize); > > > > Although this is not an align operation, but rather a "round up" > > operation. We have the NET_ROUNDUP macro already, but perhaps a new > > generic one (and then redefine the NET_ROUNDUP value to point to > > this)? > > NET_ROUNDUP is okay for this requirement and is generic. You want to make > a BLOCK_ALIGN_ROUNDUP like macro? The BlockSize variable name may vary > in different place and a new macro shall not introduce benefit. Or I > misunderstand you? The problem with NET_ROUNDUP is that it resides in NetLib.h. That's why it sounds like a generic ROUNDUP would be a useful addition to Base.h. (NET_ROUNDUP could then become an alias for that.) > > > >> + FreePages (Buffer, 1); > >> + > >> + /* Both PartitionStart and PartitionSize are counted as block size. */ > >> + Buffer = AllocatePages (EFI_SIZE_TO_PAGES (Size)); > >> + if (Buffer == NULL) { > >> + return EFI_BUFFER_TOO_SMALL; > >> + } > >> + > >> + /* Load header of boot.img */ > >> + Status = BlockIo->ReadBlocks ( > >> + BlockIo, > >> + MediaId, > >> + 0, > >> + Size, > >> + Buffer > >> + ); > >> + if (EFI_ERROR (Status)) { > >> + DEBUG ((EFI_D_ERROR, "Failed to read blocks: %r\n", Status)); > >> + goto EXIT; > >> + } > >> + > >> + Status = AbootimgBoot (Buffer, Size); > >> + > >> +EXIT: > >> + return Status; > >> +} > >> diff --git a/EmbeddedPkg/Include/Library/AbootimgLib.h b/EmbeddedPkg/Include/Library/AbootimgLib.h > >> new file mode 100644 > >> index 0000000..c0372d4 > >> --- /dev/null > >> +++ b/EmbeddedPkg/Include/Library/AbootimgLib.h > >> @@ -0,0 +1,65 @@ > >> +/** @file > >> + > >> + Copyright (c) 2013-2014, ARM Ltd. All rights reserved.
> >> + Copyright (c) 2017, Linaro. > >> + > >> + This program and the accompanying materials > >> + are licensed and made available under the terms and conditions of the BSD License > >> + which accompanies this distribution. The full text of the license may be found at > >> + http://opensource.org/licenses/bsd-license.php > >> + > >> + THE PROGRAM IS DISTRIBUTED UNDER THE BSD LICENSE ON AN "AS IS" BASIS, > >> + WITHOUT WARRANTIES OR REPRESENTATIONS OF ANY KIND, EITHER EXPRESS OR IMPLIED. > >> + > >> +**/ > >> + > >> +#ifndef __ABOOTIMG_H__ > >> +#define __ABOOTIMG_H__ > >> + > >> +#include > >> +#include > >> +#include > >> + > >> +#include > >> +#include > >> + > >> +#define BOOTIMG_KERNEL_ARGS_SIZE 512 > > > > Is this value defined somewhere? Or is it just an arbitrary size? > > It is defined by Google code. > https://android.googlesource.com/platform/system/core/+/master/mkbootimg/bootimg.h OK. It would be good to have some link in here to a more formal specification, but if one is missing, please just add this link in a comment. > > > >> + > >> +#define BOOT_MAGIC "ANDROID!" > >> +#define BOOT_MAGIC_LENGTH (sizeof (BOOT_MAGIC) - 1) > >> + > >> +/* It's the value of arm64 efi stub kernel */ > >> +#define KERNEL_IMAGE_STEXT_OFFSET 0x12C > >> +#define KERNEL_IMAGE_RAW_SIZE_OFFSET 0x130 > >> + > >> +#define FDT_SIZE_OFFSET 0x4 > > > > What are these offsets? > Seems to be offset of PE/COFF data that describe the COFF image size. > We do not need it if we do not search DTB that is attached after kernel image. OK. That would be nice if we could get rid of. If not, it would be preferable if these could be extracted via existing PE/COFF definitions. > > All of these names (BOOT/KERNEL/FDT) are a little bit to generic for > > an exported header file. Need ANDROID_BOOT_ (or similar_) prefix. > > Will do. > > > >> + > >> +typedef struct { > >> + CHAR8 BootMagic[BOOT_MAGIC_LENGTH]; > >> + UINT32 KernelSize; > >> + UINT32 KernelAddress; > >> + UINT32 RamdiskSize; > >> + UINT32 RamdiskAddress; > >> + UINT32 SecondStageBootloaderSize; > >> + UINT32 SecondStageBootloaderAddress; > >> + UINT32 KernelTaggsAddress; > >> + UINT32 PageSize; > >> + UINT32 Reserved[2]; > >> + CHAR8 ProductName[16]; > >> + CHAR8 KernelArgs[BOOTIMG_KERNEL_ARGS_SIZE]; > > > > Does the protocol not specify signedness of chars for strings? > > The BootMagic string should be "ANDROID!" to identify it is android > boot image. Sorry, what I was referring to was the use of the CHAR8 type. Checking against the link above, the original uses uint8_t, so I would have expected to see UINT8 here. Or if the original was int8_t, INT8. Since default char signedness differs between x86/ARM... > > > >> + UINT32 Id[32]; > >> +} ANDROID_BOOTIMG_HEADER; > > > > This looks identical to the definition in > > Application/AndroidFastboot/AndroidBootImg.c, only missing the > > #pragma pack(1) statement. > > > > In general there looks like a lot of code duplication between these > > two applications. Can this not be broken out into a common library > > used by both? > > Will merge the two definition. > > > >> + > >> +EFI_STATUS > >> +AbootimgGetImgSize ( > >> + IN VOID *BootImg, > >> + OUT UINTN *ImgSize > >> + ); > >> + > >> +EFI_STATUS > >> +AbootimgBoot ( > >> + IN VOID *Buffer, > >> + IN UINTN BufferSize > >> + ); > >> + > >> +#endif /* __ABOOTIMG_H__ */ > >> diff --git a/EmbeddedPkg/Library/AbootimgLib/AbootimgLib.c b/EmbeddedPkg/Library/AbootimgLib/AbootimgLib.c > >> new file mode 100644 > >> index 0000000..ea30a01 > >> --- /dev/null > >> +++ b/EmbeddedPkg/Library/AbootimgLib/AbootimgLib.c > >> @@ -0,0 +1,350 @@ > >> +/** @file > >> + > >> + Copyright (c) 2013-2014, ARM Ltd. All rights reserved.
> >> + Copyright (c) 2017, Linaro. All rights reserved. > >> + > >> + This program and the accompanying materials > >> + are licensed and made available under the terms and conditions of the BSD License > >> + which accompanies this distribution. The full text of the license may be found at > >> + http://opensource.org/licenses/bsd-license.php > >> + > >> + THE PROGRAM IS DISTRIBUTED UNDER THE BSD LICENSE ON AN "AS IS" BASIS, > >> + WITHOUT WARRANTIES OR REPRESENTATIONS OF ANY KIND, EITHER EXPRESS OR IMPLIED. > >> + > >> +**/ > >> + > >> +#include > >> +#include > >> +#include > >> +#include > >> + > >> +#include > >> +#include > >> + > >> +#include > >> + > >> +// Check Val (unsigned) is a power of 2 (has only one bit set) > >> +#define IS_POWER_OF_2(Val) (Val != 0 && ((Val & (Val - 1)) == 0)) > > > > Missing parentheses in macro. > > Will fix. > > > >> + > >> +typedef struct { > >> + MEMMAP_DEVICE_PATH Node1; > >> + EFI_DEVICE_PATH_PROTOCOL End; > >> +} MEMORY_DEVICE_PATH; > >> + > >> +STATIC ABOOTIMG_PROTOCOL *mAbootimg; > >> + > >> +STATIC CONST MEMORY_DEVICE_PATH MemoryDevicePathTemplate = > >> +{ > >> + { > >> + { > >> + HARDWARE_DEVICE_PATH, > >> + HW_MEMMAP_DP, > >> + { > >> + (UINT8)(sizeof (MEMMAP_DEVICE_PATH)), > >> + (UINT8)((sizeof (MEMMAP_DEVICE_PATH)) >> 8), > >> + }, > >> + }, // Header > >> + 0, // StartingAddress (set at runtime) > >> + 0 // EndingAddress (set at runtime) > >> + }, // Node1 > >> + { > >> + END_DEVICE_PATH_TYPE, > >> + END_ENTIRE_DEVICE_PATH_SUBTYPE, > >> + { sizeof (EFI_DEVICE_PATH_PROTOCOL), 0 } > >> + } // End > >> +}; > >> + > >> +EFI_STATUS > >> +AbootimgGetImgSize ( > >> + IN VOID *BootImg, > >> + OUT UINTN *ImgSize > >> + ) > >> +{ > >> + ANDROID_BOOTIMG_HEADER *Header; > >> + > >> + Header = (ANDROID_BOOTIMG_HEADER *) BootImg; > >> + > >> + if (AsciiStrnCmp (Header->BootMagic, BOOT_MAGIC, BOOT_MAGIC_LENGTH) != 0) { > >> + return EFI_INVALID_PARAMETER; > >> + } > >> + > >> + ASSERT (IS_POWER_OF_2 (Header->PageSize)); > > > > I am not convinced this is the right thing to check for. > > Not every power of two is a valid page size. > > Semantically, any macro here should really be called > > IS_VALID_PAGE_SIZE(). > > This page_size is not related to UEFI directly because it is defined > by Android code. It is just > the size of boot header in boot image. The kernel image start from > offset of page_size. You can > check the comment in the android boot code link. Right, OK. That still sort of reinforces my view that it being a power of 2 is not really adding more robustness. By all means check if the field has actually been initialized, but beyond that, surely we need to expect that its value is relevant to the rest of the contents of the image? Or is this more a safety check based on arithmetic elsewhere depending on it being a power of 2? > >> + > >> + /* Get real size of abootimg */ > >> + *ImgSize = ALIGN_VALUE (Header->KernelSize, Header->PageSize) + > >> + ALIGN_VALUE (Header->RamdiskSize, Header->PageSize) + > >> + ALIGN_VALUE (Header->SecondStageBootloaderSize, Header->PageSize) + > >> + Header->PageSize; > >> + return EFI_SUCCESS; > >> +} > >> + > >> +EFI_STATUS > >> +AbootimgGetKernelInfo ( > >> + IN VOID *BootImg, > >> + OUT VOID **Kernel, > >> + OUT UINTN *KernelSize > >> + ) > >> +{ > >> + ANDROID_BOOTIMG_HEADER *Header; > >> + > >> + Header = (ANDROID_BOOTIMG_HEADER *) BootImg; > >> + > >> + if (AsciiStrnCmp (Header->BootMagic, BOOT_MAGIC, BOOT_MAGIC_LENGTH) != 0) { > >> + return EFI_INVALID_PARAMETER; > >> + } > >> + > >> + if (Header->KernelSize == 0) { > >> + return EFI_NOT_FOUND; > >> + } > >> + > >> + ASSERT (IS_POWER_OF_2 (Header->PageSize)); > >> + > >> + *KernelSize = Header->KernelSize; > >> + *Kernel = BootImg + Header->PageSize; > >> + return EFI_SUCCESS; > >> +} > >> + > >> +EFI_STATUS > >> +AbootimgGetRamdiskInfo ( > >> + IN VOID *BootImg, > >> + OUT VOID **Ramdisk, > >> + OUT UINTN *RamdiskSize > >> + ) > >> +{ > >> + ANDROID_BOOTIMG_HEADER *Header; > >> + UINT8 *BootImgBytePtr; > >> + > >> + // Cast to UINT8 so we can do pointer arithmetic > >> + BootImgBytePtr = (UINT8 *) BootImg; > >> + > >> + Header = (ANDROID_BOOTIMG_HEADER *) BootImg; > >> + > >> + if (AsciiStrnCmp (Header->BootMagic, BOOT_MAGIC, BOOT_MAGIC_LENGTH) != 0) { > >> + return EFI_INVALID_PARAMETER; > >> + } > >> + > >> + ASSERT (IS_POWER_OF_2 (Header->PageSize)); > >> + > >> + *RamdiskSize = Header->RamdiskSize; > >> + > >> + if (Header->RamdiskSize != 0) { > >> + *Ramdisk = (VOID *) (BootImgBytePtr > >> + + Header->PageSize > >> + + ALIGN_VALUE (Header->KernelSize, Header->PageSize)); > >> + } > >> + return EFI_SUCCESS; > >> +} > >> + > >> +EFI_STATUS > >> +AbootimgGetSecondBootLoaderInfo ( > >> + IN VOID *BootImg, > >> + OUT VOID **Second, > >> + OUT UINTN *SecondSize > >> + ) > >> +{ > >> + ANDROID_BOOTIMG_HEADER *Header; > >> + UINT8 *BootImgBytePtr; > >> + > >> + // Cast to UINT8 so we can do pointer arithmetic > >> + BootImgBytePtr = (UINT8 *) BootImg; > >> + > >> + Header = (ANDROID_BOOTIMG_HEADER *) BootImg; > >> + > >> + if (AsciiStrnCmp (Header->BootMagic, BOOT_MAGIC, BOOT_MAGIC_LENGTH) != 0) { > >> + return EFI_INVALID_PARAMETER; > >> + } > >> + > >> + ASSERT (IS_POWER_OF_2 (Header->PageSize)); > >> + > >> + *SecondSize = Header->SecondStageBootloaderSize; > >> + > >> + if (Header->SecondStageBootloaderSize != 0) { > >> + *Second = (VOID *) (BootImgBytePtr > >> + + Header->PageSize > >> + + ALIGN_VALUE (Header->KernelSize, Header->PageSize) > >> + + ALIGN_VALUE (Header->RamdiskSize, Header->PageSize)); > >> + } > >> + return EFI_SUCCESS; > >> +} > >> + > >> +EFI_STATUS > >> +AbootimgGetKernelArgs ( > >> + IN VOID *BootImg, > >> + OUT CHAR8 *KernelArgs > >> + ) > >> +{ > >> + ANDROID_BOOTIMG_HEADER *Header; > >> + > >> + Header = (ANDROID_BOOTIMG_HEADER *) BootImg; > >> + AsciiStrnCpyS (KernelArgs, BOOTIMG_KERNEL_ARGS_SIZE, Header->KernelArgs, > >> + BOOTIMG_KERNEL_ARGS_SIZE); > >> + > >> + return EFI_SUCCESS; > >> +} > >> + > >> +EFI_STATUS > >> +AbootimgInstallFdt ( > >> + IN VOID *BootImg, > >> + IN EFI_PHYSICAL_ADDRESS FdtBase, > >> + OUT VOID *KernelArgs > >> + ) > >> +{ > >> + VOID *Ramdisk; > >> + UINTN RamdiskSize; > >> + CHAR8 ImgKernelArgs[BOOTIMG_KERNEL_ARGS_SIZE]; > >> + INTN err; > >> + EFI_STATUS Status; > >> + EFI_PHYSICAL_ADDRESS NewFdtBase; > >> + > >> + Status = gBS->LocateProtocol (&gAbootimgProtocolGuid, NULL, (VOID **) &mAbootimg); > >> + if (EFI_ERROR (Status)) { > >> + return Status; > >> + } > >> + > >> + if (EFI_ERROR (Status)) { > >> + return Status; > >> + } > >> + > >> + Status = AbootimgGetRamdiskInfo ( > >> + BootImg, > >> + &Ramdisk, > >> + &RamdiskSize > >> + ); > >> + if (EFI_ERROR (Status)) { > >> + return Status; > >> + } > >> + Status = AbootimgGetKernelArgs ( > >> + BootImg, > >> + ImgKernelArgs > >> + ); > >> + if (EFI_ERROR (Status)) { > >> + return Status; > >> + } > >> + // Get kernel arguments from Android boot image > >> + AsciiStrToUnicodeStrS (ImgKernelArgs, KernelArgs, BOOTIMG_KERNEL_ARGS_SIZE >> 1); > >> + // Set the ramdisk in command line arguments > >> + UnicodeSPrint ( > >> + (CHAR16 *)KernelArgs + StrLen (KernelArgs), BOOTIMG_KERNEL_ARGS_SIZE, > >> + L" initrd=0x%x,0x%x", > >> + (UINTN)Ramdisk, (UINTN)RamdiskSize > >> + ); > > > > This is not an appropriate way to set ramdisk. There are dedicated > > "linux,initrd-start" and "linux,initrd-end" options in the chosen node. > > See > > http://git.savannah.gnu.org/cgit/grub.git/tree/grub-core/loader/arm64/linux.c#n79 > > for an example. > > Will try this one. > > > >> + > >> + // Append platform kernel arguments > >> + Status = mAbootimg->AppendArgs (KernelArgs, BOOTIMG_KERNEL_ARGS_SIZE); > > > > Am I missing something? Where is this AppendArgs function defined? > > The function body is implemented in platform driver. We can check the > NULL pointer before call it. OK. It would be nice to see the code being upstreamed including code that actually uses it. Otherwise, I am only reviewing one side of the interface. > > > >> + if (EFI_ERROR (Status)) { > >> + return Status; > >> + } > >> + > >> + Status = mAbootimg->UpdateDtb (FdtBase, &NewFdtBase); > > > > Likewise, where is this function defined? > implemented in platform driver too. > > > > >> + if (EFI_ERROR (Status)) { > >> + return Status; > >> + } > >> + > >> + // > >> + // Sanity checks on the new FDT blob. > >> + // > >> + err = fdt_check_header ((VOID*)(UINTN)NewFdtBase); > >> + if (err != 0) { > >> + Print (L"ERROR: Device Tree header not valid (err:%d)\n", err); > >> + return EFI_INVALID_PARAMETER; > >> + } > >> + > >> + Status = gBS->InstallConfigurationTable ( > >> + &gFdtTableGuid, > >> + (VOID *)(UINTN)NewFdtBase > >> + ); > >> + return Status; > >> +} > >> + > >> +EFI_STATUS > >> +AbootimgBoot ( > >> + IN VOID *Buffer, > >> + IN UINTN BufferSize > >> + ) > >> +{ > >> + EFI_STATUS Status; > >> + VOID *Kernel; > >> + UINTN KernelSize; > >> + VOID *SecondLoader; > >> + UINTN SecondLoaderSize; > >> + MEMORY_DEVICE_PATH KernelDevicePath; > >> + EFI_HANDLE ImageHandle; > >> + EFI_PHYSICAL_ADDRESS FdtBase; > >> + VOID *NewKernelArg; > >> + EFI_LOADED_IMAGE_PROTOCOL *ImageInfo; > >> + INTN Err; > >> + > >> + Status = AbootimgGetKernelInfo ( > >> + Buffer, > >> + &Kernel, > >> + &KernelSize > >> + ); > >> + if (EFI_ERROR (Status)) { > >> + return Status; > >> + } > >> + > >> + /* For flatten image, Fdt is attached at the end of kernel. > >> + Get real kernel size. > >> + */ > > > > This is a completely broken usage model :( > > Is this a requirement for this support? > > > > I mean, the solution you have implemented is a nice way of dealing > > with the mess, but per-kernel-image DT blobs were deprecated at kernel > > summit 2013. > > This is to support legacy hikey bootimage. We can drop this support if > Haojian does not object. I will leave that to him to answer. > >> + KernelSize = *(UINT32 *)((EFI_PHYSICAL_ADDRESS)(UINTN)Kernel + KERNEL_IMAGE_STEXT_OFFSET) + > >> + *(UINT32 *)((EFI_PHYSICAL_ADDRESS)(UINTN)Kernel + KERNEL_IMAGE_RAW_SIZE_OFFSET); > >> + > >> + NewKernelArg = AllocateZeroPool (BOOTIMG_KERNEL_ARGS_SIZE); > >> + if (NewKernelArg == NULL) { > >> + DEBUG ((DEBUG_ERROR, "Fail to allocate memory\n")); > >> + return EFI_OUT_OF_RESOURCES; > >> + } > >> + > >> + /* FDT is at the end of kernel image */ > >> + FdtBase = (EFI_PHYSICAL_ADDRESS)(UINTN)Kernel + KernelSize; > >> + // > >> + // Sanity checks on the original FDT blob. > >> + // > >> + Err = fdt_check_header ((VOID*)(UINTN)FdtBase); > >> + if (Err != 0) { > >> + /* Check whether FDT is located in second boot loader */ > > > > What is a second boot loader? > > Per description in code, it is the second stage boot loader that come first > than kernel. But it seems that no platform adopt this boot loader usage > and second_size is commonly 0. On the contrary, there are cases to put > DTB in the second stage as far as I know. UEFI android boot app parse > kernel directly Yes, I would have hoped so. > and second stage surely can be occupied by DTB. This does get semantically dubious. But if it is hidden away in a helper function, and well commented, then I guess this is OK. > >> + Status = AbootimgGetSecondBootLoaderInfo ( > >> + Buffer, > >> + &SecondLoader, > >> + &SecondLoaderSize > >> + ); > >> + if (EFI_ERROR (Status)) { > >> + return Status; > >> + } > >> + > >> + Err = fdt_check_header ((VOID*)(UINTN)SecondLoader); > >> + if (Err != 0) { > >> + DEBUG ((DEBUG_ERROR, "ERROR: Device Tree header not valid (Err:%d)\n", Err)); > >> + return EFI_INVALID_PARAMETER; > >> + } > >> + FdtBase = (EFI_PHYSICAL_ADDRESS)SecondLoader; > >> + } > >> + > > > > The above two scenarios could do with broken out as separate helper > > functions. Preferably also with a Pcd to determine whether the > > application should even bother to go looking if a platform-provided > > device tree was already installed. > > Could you help elaborate more on how to check DTB installation status? > This android boot app will read block device for kernel/dtb and cmd lines > directly as the first place. So I do not know what code will install the DTB > before android boot image is read. Use EfiGetSystemConfigurationTable to attempt to retrieve the guid you are currently installing here. Tying kernel image and DTB together is a travesty. I know some people just can't get past the point, so we need to enable them to keep doing it, but the expected behaviour is that the platform initialisation code registers a table (whatever its origin). > >> + Status = AbootimgInstallFdt (Buffer, FdtBase, NewKernelArg); > >> + if (EFI_ERROR (Status)) { > >> + FreePool (NewKernelArg); > >> + return EFI_INVALID_PARAMETER; > >> + } > >> + > >> + KernelDevicePath = MemoryDevicePathTemplate; > >> + > >> + // Have to cast to UINTN before casting to EFI_PHYSICAL_ADDRESS in order to > >> + // appease GCC. > > > > (Redundant comment. It's a common enough problem.) > > Will remove. > > > >> + KernelDevicePath.Node1.StartingAddress = (EFI_PHYSICAL_ADDRESS)(UINTN) Kernel; > >> + KernelDevicePath.Node1.EndingAddress = (EFI_PHYSICAL_ADDRESS)(UINTN) Kernel + KernelSize; > >> + > >> + Status = gBS->LoadImage (TRUE, gImageHandle, (EFI_DEVICE_PATH *)&KernelDevicePath, (VOID*)(UINTN)Kernel, KernelSize, &ImageHandle); > > > > Please split function call up over multiple lines. > > Will do. > > > >> + > >> + // Set kernel arguments > >> + Status = gBS->HandleProtocol (ImageHandle, &gEfiLoadedImageProtocolGuid, (VOID **) &ImageInfo); > > > > Please split function call up over multiple lines. > > Will do. > > > >> + ImageInfo->LoadOptions = NewKernelArg; > >> + ImageInfo->LoadOptionsSize = StrLen (NewKernelArg) * sizeof (CHAR16); > >> + > >> + // Before calling the image, enable the Watchdog Timer for the 5 Minute period > > > > Please wrap comment at 80 characters. (Or delete the "for the 5 Minute > > period bit based on the next comment.) > > Will do. > > > >> + gBS->SetWatchdogTimer (5 * 60, 0x0000, 0x00, NULL); > > > > Please create a define for that timeout. ANDROID_BOOT_WATCHDOG_TIMEOUT > > or something. > > Or, hmm, given that the UEFI specification (2.7) says "The watchdog > > must be set to a period of 5 minutes", this may even be worth putting > > into MdePkg/MdeModulePkg under a generic name. > > You want to add below service call usage? I prefer a definition of max timeout. > We can make another patch to change all max timeout usage later. > gBS->SetWatchdogMaxTimeout (0x0000, 0x00, NULL); > #define MAX_WATCHDOG_TIMEOUT (60*5) Yes, that works. > > > > The UEFI specification says "The firmware reserves codes 0x0000 to > > 0xFFFF. Loaders and operating systems may use other timeout codes.". > > So I think this code needs to set this to something else. > > You are right. 0x0000 should be a copy/past bug here. Will change to 0x10000 > as fastboot app. > > > > You don't need to say 0x00 when you mean 0. > > > >> + // Start the image > >> + Status = gBS->StartImage (ImageHandle, NULL, NULL); > >> + // Clear the Watchdog Timer after the image returns > > > > _if_ the image returns, surely? > You mean it never return? If coff data of kernel image is corrupted, > it surely return. Of course, but that is the error condition - not the expected behaviour. So it is "if", not "when" (which implies inevitability). > > > >> + gBS->SetWatchdogTimer (0x0000, 0x0000, 0x0000, NULL); > > > > Don't say 0x* when you mean 0. > Will change it. > > > >> + return EFI_SUCCESS; > >> +} > >> diff --git a/EmbeddedPkg/Library/AbootimgLib/AbootimgLib.inf b/EmbeddedPkg/Library/AbootimgLib/AbootimgLib.inf > >> new file mode 100644 > >> index 0000000..461dcb8 > >> --- /dev/null > >> +++ b/EmbeddedPkg/Library/AbootimgLib/AbootimgLib.inf > >> @@ -0,0 +1,48 @@ > >> +#/** @file > >> +# > >> +# Copyright (c) 2013-2015, ARM Ltd. All rights reserved.
> >> +# Copyright (c) 2017, Linaro. All rights reserved. > >> +# > >> +# This program and the accompanying materials > >> +# are licensed and made available under the terms and conditions of the BSD License > >> +# which accompanies this distribution. The full text of the license may be found at > >> +# http://opensource.org/licenses/bsd-license.php > >> +# THE PROGRAM IS DISTRIBUTED UNDER THE BSD LICENSE ON AN "AS IS" BASIS, > >> +# WITHOUT WARRANTIES OR REPRESENTATIONS OF ANY KIND, EITHER EXPRESS OR IMPLIED. > >> +# > >> +# > >> +#**/ > >> + > >> +[Defines] > >> + INF_VERSION = 0x00010019 > >> + BASE_NAME = AbootimgLib > >> + FILE_GUID = ed3b8739-6fa7-4cb1-8aeb-2496f8fcaefa > >> + MODULE_TYPE = BASE > >> + VERSION_STRING = 1.0 > >> + LIBRARY_CLASS = AbootimgLib > >> + > >> +# > >> +# The following information is for reference only and not required by the build tools. > >> +# > >> +# VALID_ARCHITECTURES = ARM AARCH64 > >> +# > >> + > >> +[Sources] > >> + AbootimgLib.c > >> + > >> +[LibraryClasses] > >> + DebugLib > >> + FdtLib > >> + PrintLib > >> + UefiBootServicesTableLib > >> + UefiLib > >> + > >> +[Packages] > >> + MdePkg/MdePkg.dec > >> + EmbeddedPkg/EmbeddedPkg.dec > > > > Please sort alphabetically. > > Will do. > > > > / > > Leif > > > >> + > >> +[Protocols] > >> + gAbootimgProtocolGuid > >> + > >> +[Guids] > >> + gFdtTableGuid > >> -- > >> 1.9.1 > >> > > Thanks for review with so much details. > > Jun