[External] Re: Clarify some questions about in-band firmware update

Sat Jun 29 00:09:55 AEST 2019

On Fri, Jun 28, 2019 at 4:30 AM Andrew MS1 Peng <pengms1 at lenovo.com> wrote:

> Hi Patrick,
>
>
> I synchronized our Openbmc repository with *upstream* repository then
> compiled and checked the result yesterday, we can generate *dynamically
> linked binary but still failed to generate statically linked binary.  *I
> used *OpenBmc *environment to compile and also created a SDK environment
> by using x86-64 toolchain (oecore-x86_64-core2-64-toolchain-nodistro.0.sh)
> to compile, the compile result is the same. *It seems to the linker not
> find the *archive file libpci.a, *could you take a look my problem?
> Thanks.*
>
>
>
> *I will get error message when I run the tool in my host platform as
> below. The** dynamically linked binary** is not convenience to use since
> I have to modify *rpath and interpreter of already compiled ELF.
>
> pengms1 at hsbmc: $ ./burn_my_bmc
>
> -bash: ./burn_my_bmc: No such file or directory
>
> pengms1 at hsbmc: $ ldd ./burn_my_bmc
> *./burn_my_bmc: /lib/x86_64-linux-gnu/libpci.so.3: version `LIBPCI_3.5'
> not found (required by ./burn_my_bmc)*
>
> *       linux-vdso.so.1 =>  (0x00007ffcd2363000)*
>
> *       libipmiblob.so.0 =>
> /home/pengms1/SDK/OpenBmc/rootfs/lib/libipmiblob.so.0 (0x00007f25d373a000)*
>
> *       libpci.so.3 => /lib/x86_64-linux-gnu/libpci.so.3
> (0x00007f25d3512000)*
>
> *       libstdc++.so.6 => /usr/lib/x86_64-linux-gnu/libstdc++.so.6
> (0x00007f25d3190000)*
>
> *       libgcc_s.so.1 => /lib/x86_64-linux-gnu/libgcc_s.so.1
> (0x00007f25d2f79000)*
>
> *       libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007f25d2baf000)*
>
> *       libm.so.6 => /lib/x86_64-linux-gnu/libm.so.6 (0x00007f25d28a6000)*
>
> *       libz.so.1 => /lib/x86_64-linux-gnu/libz.so.1 (0x00007f25d268b000)*
>
> *       libresolv.so.2 => /lib/x86_64-linux-gnu/libresolv.so.2
> (0x00007f25d2470000)*
>
> *       libudev.so.1 => /lib/x86_64-linux-gnu/libudev.so.1
> (0x00007f25d2450000)*
>
> *       /lib/ld-linux-x86-64.so.2 => /lib64/ld-linux-x86-64.so.2
> (0x000055731193b000)*
>
> *       librt.so.1 => /lib/x86_64-linux-gnu/librt.so.1
> (0x00007f25d2247000)*
>
> *       libpthread.so.0 => /lib/x86_64-linux-gnu/libpthread.so.0
> (0x00007f25d202a000)*
>
>
>
> *The compile error message is below:*
>
> x86_64-openbmc-linux-libtool: link: x86_64-openbmc-linux-g++ -m64
> -march=core2 -mtune=core2 -msse3 -mfpmath=sse
> --sysroot=/home/pengms1/openbmc_trunk_x86/build/tmp/work/core2-64-openbmc-linux/phosphor-ipmi-flash/0.1+gitAUTOINC+7c79b252c1-r1/recipe-sysroot
> -std=c++17 -I../../git -O2 -pipe -g -feliminate-unused-debug-types
> -fmacro-prefix-map=/home/pengms1/openbmc_trunk_x86/build/tmp/work/core2-64-openbmc-linux/phosphor-ipmi-flash/0.1+gitAUTOINC+7c79b252c1-r1=/usr/src/debug/phosphor-ipmi-flash/0.1+gitAUTOINC+7c79b252c1-r1
> -fdebug-prefix-map=/home/pengms1/openbmc_trunk_x86/build/tmp/work/core2-64-openbmc-linux/phosphor-ipmi-flash/0.1+gitAUTOINC+7c79b252c1-r1=/usr/src/debug/phosphor-ipmi-flash/0.1+gitAUTOINC+7c79b252c1-r1
> -fdebug-prefix-map=/home/pengms1/openbmc_trunk_x86/build/tmp/work/core2-64-openbmc-linux/phosphor-ipmi-flash/0.1+gitAUTOINC+7c79b252c1-r1/recipe-sysroot=
> -fdebug-prefix-map=/home/pengms1/openbmc_trunk_x86/build/tmp/work/core2-64-openbmc-linux/phosphor-ipmi-flash/0.1+gitAUTOINC+7c79b252c1-r1/recipe-sysroot-native=
> -fvisibility-inlines-hidden -Wall -Werror -DENABLE_REBOOT_UPDATE
> -DENABLE_STATIC_LAYOUT -DENABLE_LPC_BRIDGE -DASPEED_LPC -static -o
> burn_my_bmc burn_my_bmc-main.o  ./.libs/libupdater.a -lipmiblob -lpci
>
> |
> /home/pengms1/openbmc_trunk_x86/build/tmp/work/core2-64-openbmc-linux/phosphor-ipmi-flash/0.1+gitAUTOINC+7c79b252c1-r1/recipe-sysroot-native/usr/bin/x86_64-openbmc-linux/../../libexec/x86_64-openbmc-linux/gcc/x86_64-openbmc-linux/9.1.0/ld:
> *cannot find -lpci*
>

That's indicating that it couldn't find the pciutils installed.  Like I
said yesterday, I hadn't tried to build the host-tool in this environment
and had to clean up the dependency checks for it.  That said, I went
through and fixed it all and got it to build as expected from a pure ubuntu
docker container.  Those patches are up for review and not yet in OpenBMC.
That said, I dont' know why the lpci might not be found for that
compilation other than possibly the pciutils recipe is broken.  I found
yesterday while getting this to work that the instructions on the pciutils
README weren't quite right for building the library AND installing that
library.

So I wrote up custom instructions in the README for phosphor-ipmi-flash,
that patch is under review.

If you want to debut your current environment, you need to see if you can
find the pkgconfig file and libraries defined:
https://github.com/pciutils/pciutils/blob/master/lib/libpci.pc.in

Basically search for libpci.pc and then read that file.  Verify the shared
library objects are in the SDK environment.

If you need the libpci archive object for static compilation, that may not
be installed by default, you'll want to read the pciutils README.

>
>
> *My OpenBmc build **environment is below.*
>
> OpenBmc Setting
>
> phosphor-ipmi-flash setting (build *static binary*)
>
> Extra setting
>
> *Build Configuration:*
>
> *BB_VERSION           = "1.43.0"*
>
> *BUILD_SYS            = "x86_64-linux"*
>
> *NATIVELSBSTRING      = "ubuntu-16.04"*
>
> *TARGET_SYS           = "x86_64-openbmc-linux"*
>
> *MACHINE              = "qemux86-64"*
>
> *DISTRO               = "openbmc-phosphor"*
>
> *DISTRO_VERSION       = "0.1.2"*
>
> *TUNE_FEATURES        = "m64 core2"*
>
> *TARGET_FPU           = ""*
>
> EXTRA_OECONF += " --enable-build-host-tool --enable-static-layout
> --enable-lpc-bridge
> --enable-aspeed-lpc MAPPED_ADDRESS=0x98000000 --enable-reboot-update "
>
> EXTRA_OEMAKE += " LDFLAGS=-all-static " (Build a *statically linked
> binary*)
>
> DISABLE_STATIC = ""  (Build *statically library for *ipmi-blob-tool)
>
> phosphor-ipmi-flash setting (build *dynamic binary*)
>
> EXTRA_OECONF += " --enable-build-host-tool --enable-static-layout
> --enable-lpc-bridge
> --enable-aspeed-lpc MAPPED_ADDRESS=0x98000000 --enable-reboot-update "
>
>
>
> Regards,
> Andrew
>
>
>
> *发件人:* Patrick Venture <venture at google.com>
> *发送时间:* 2019年6月27日 9:59
> *收件人:* Andrew MS1 Peng <pengms1 at lenovo.com>; OpenBMC Maillist <
> openbmc at lists.ozlabs.org>
> *抄送:* Duke KH Du <dukh at lenovo.com>; benjaminfair at google.com; Adriana
> Kobylak <anoo at us.ibm.com>
> *主题:* [External] Re: Clarify some questions about in-band firmware update
>
>
>
>
>
>
>
> On Wed, Jun 26, 2019 at 7:44 AM Patrick Venture <venture at google.com>
> wrote:
>
>
>
>
>
> On Wed, Jun 26, 2019 at 4:28 AM Andrew MS1 Peng <pengms1 at lenovo.com>
> wrote:
>
> Hi Patrick,
>
>
>
> I have some questions as below, could you help to clarify. Thanks.
>
> 1.      Do you have any plans to support BIOS update in
> phosphor-ipmi-flash?
>
> If the answer is no, we would like to help to add this feature. Could you
> give us some advice or clues to implement it.
>
>
>
> This could be added to the design (first step).
>
> I could probably roll out this in a day or two, it's not a difficult
> extension with a few caveats. We'd have to define new services, such as
> update_bios.service instead of update_bmc.service.
>
>
>
> I have some cycles, so I don't mind reviewing the changes required and
> rolling something out --
>
>
>
> 2.      The FPGA rom image is embedded in BMC image, we would like to add
> fpga-update.service and fpga-verify.service in phosphor-ipmi-flash, we
> expect that run both services after update BMC image completely. Do you
> have any concerns about it?
>
>
>
> There are a couple things to unpack here.  The FPGA rom is embedded in the
> BMC image.  How is the BMC updated?  I ask because a machine can (and is
> expected to) provide its own update_bmc.service implementation.  This
> service could trigger the fpga-verify and fpga-update services within
> whatever code is called from the update_bmc service you specify.
>
>
>
> 3.      How can I make a static executable host tool (burn_my_bmc)?
> Currently, I can’t to make a *statically linked binary successfully*, but
> I have made it before you added p2a function.
>
> For the host tool you need to specify very little, you should be able to
> build it via:
>
> ./bootstrap.sh
>
> ./configure
>
> ./make
>
>
>
> You will need to checkout and install ipmi-blob-tool first
> (openbmc/ipmi-blob-tool)
>
>
>
>
>
> ^^ --- those steps should go on the README.  Will add them shortly.
>
>
>
> Based on different OS environment as below, I would like to have a static
> executable since *dynamically linked binary can’t work normally in host
> OS.*
>
> Host OS environment
>
> Build environment (OpenBmc)
>
> Gcc  v4.8.5
>
> Gcc   v8.3.0
>
> Glibc  v2.17
>
> Glibc  v2.29
>
>
>
> The bitbake recipe isn't meant to be used for the host tool.  It
> explicitly disables building the host-tool by default.  You should be able
> to check out the repo and build from there.  Let me know if you run into
> issues with that.
>
>
>
> 4.      I didn’t find the below service files in source tree, where can
> find default services or should we provide our own services?
>
> update_bmc.service
> verify_image.service
> prepare_update.service
>
>
>
> That's correct.  There are currently no default service files.  I am
> thinking of providing a default prepare_update service in a later CL,
> however updating the BMC or verifying the image depend entirely on the
> system -- which makes it difficult to have a default.  In point of fact,
> there isn't really a valid default phosphor-ipmi-flash build for a BMC.
> Everything needs to be specified.
>
>
>
> ^--- I can work on making the ubi bmc option available if the ubi distro
> feature is enabled, etc -- providing some defaults.
>
> ^--- I have a plan to provide a prepare_update.service once I finish
> testing Aspeed LPC support.
>
>
>
>
>
> I've been reviewing the build process for the host-tool and the configure
> file and other pieces require the BMC-stuff to be installed first
> (sdbusplus, etc).  I'm going to cleanup the configure file so that if
> you're building the host-tool it doesn't try to build the BMC stuff.
>
>
>
>
>
> Regards,
> Andrew
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ozlabs.org/pipermail/openbmc/attachments/20190628/c3c9d9f2/attachment-0001.htm>