Re: [edk2-devel] Problem with decompression on EDK2

["Laszlo Ersek" <lersek@redhat.com>]

On 09/26/19 11:27, Loh, Tien Hock wrote:
> Hi
> 
> I have an issue while porting a new platform to EDK2. After porting the Stratix 10 platform (ARM) to EDK2 it works correctly. However, when I tried to add more INF to the FDF file, it failed to decompress the image during boot time, log as below
> INFO:    DDR: DRAM calibration success.
> INFO:    Scrubbing ECC
> INFO:    Init HPS NOC's DDR Scheduler.
> NOTICE:  BL2: v2.1(debug):v2.1-15-g5880144-dirty
> NOTICE:  BL2: Built : 14:02:01, Aug 22 2019
> INFO:    BL2: Doing platform setup
> INFO:    BL2: Loading image id 3
> INFO:    Loading image id=3 at address 0xffe1c000
> INFO:    Image id=3 loaded: 0xffe1c000 - 0xffe22019
> INFO:    BL2: Loading image id 5
> INFO:    Loading image id=5 at address 0x50000
> INFO:    Image id=5 loaded: 0x50000 - 0x150000
> NOTICE:  BL2: Booting BL31
> INFO:    Entry point address = 0xffe1c000
> INFO:    SPSR = 0x3cd
> NOTICE:  BL31: v2.1(release):v2.1-604-g3441952
> NOTICE:  BL31: Built : 15:32:55, Sep 25 2019
> UEFI firmware (version 1.0 built at 15:15:26 on Sep 26 2019)
> Decompress Failed - Invalid Parameter
> 
> ASSERT_EFI_ERROR (Status = Not Found)
> ASSERT [ArmPlatformPrePiUniCore] /nfs/png/disks/swuser_work_thloh/push/build/edk2/ArmPlatformPkg/PrePi/PrePi.c(151): !EFI_ERROR (Status)
> 
> This happens if I add more items to the image, and if I reduce the size of the image, it works correctly again. This could be a clue to what could've gone wrong. Does anyone have any ideas what might've gone wrong?
> I traced the decompression to LzmaDecode, and got lost in the code there.

The compressed section in a firmware volume is likely truncated at build
time, and does not decompress at boot time. I would expect the "build"
utility to report this issue, and stop with an error.

When you run "build" in the successful and in the failing cases, what
are the stats, respectively, that are printed at the end of the build logs?

Also, you could check your FDF file(s) near
"EE4E5898-3914-4259-9D6E-DC7BD79403CF"; perhaps you find interesting bits.

Furthermore, if the boundary seems to be 16MB, then it could be an issue
in the processing code. See "ExtendedSize" in
"MdePkg/Include/Pi/PiFirmwareFile.h". I vaguely recall issues around
some code trying to parse EFI_FFS_FILE_HEADER2 as EFI_FFS_FILE_HEADER,
or EFI_COMMON_SECTION_HEADER2 as EFI_COMMON_SECTION_HEADER.

Thanks
Laszlo