[PATCHv5 00/12] kexec: Use BPF lskel to enable kexec to load PE format boot image
Philipp Rudo
prudo at redhat.com
Thu Sep 18 06:43:11 PDT 2025
Hi Pingfan,
On Tue, 16 Sep 2025 10:00:41 +0800
Pingfan Liu <piliu at redhat.com> wrote:
> On Mon, Sep 01, 2025 at 04:29:29PM +0200, Philipp Rudo wrote:
> Hi Philipp,
>
> Thank you for deep insight, please see the comments
>
> > Hi Pingfan,
> >
> > thanks for sharing the updated version of the series. There are a few
> > small nits you can find in my comments to the individual patches.
> >
> > I also took an other look at the bigger picture. The way I see it the
> > series contains two major changes.
> >
> > 1. A generic mechanism to parse and run bpf programs during kexec.
>
> Thanks for your suggestion. This makes the whole design look more
> natural. Without it, using the PE format parser as a frontend to the
> image parser looks a bit awkward.
>
> > 2. A new loader for UEFI Applications building up on 1.
> >
> > Both those changes are currently smashed together as "PE image loader",
> > which IMHO is quite confusing. It's correct that UEFI Apps are PE
> > files. But the PE format is also used in many other ways and in the end
> > we are only interested in this specific use case. Plus the generic
> > mechanism to parse and run bpf programs during kexec can also be used
> > with any other file format, not just PE.
> >
> > In addition I noticed that hooking into kexec_file while loading the
> > image is too late. The problem is that in
> > kernel/kexec_file.c:kimage_file_prepare_segments the cmdline is
> > measured for IMA before the image is loaded. But we allow the
> > bpf_prog to change the cmdline. When that is done the IMA measurement
> > no longer is correct. So we need a new hook to run the bpf programs
> > after the initrd and cmdline were read but before the IMA measurement is
> > done.
> >
>
> This should be a issue when turning on IMA measurement on cmdline. And it
> can be perfectly addressed with your suggestion.
>
> > So my suggestions are:
> >
> > 1. Extend the kexec_file_ops by a new 'get_bpf_prog' hook.
>
> I'm not sure I agree with you on this. We both agree that the generic
> mechanism to parse and run BPF programs should be independent of the
> image parser and belong in kexec_file.c. However, kexec_file_ops is
> still a concept that belongs to the image parser layer. So I think that
> moving the logic to kexec_file.c is good enough.
I'm not sure I get your point here. My rational is, that the bpf_prog
can be included in different file formats. So we still need a mechanism
to parse it from the different formats. Plus we need a file_ops that is
"responsible" for a file type. All in all it makes most sense to me to
have a file_ops that only has a .probe and .get_bpf_prog.
But yes, maybe I'm missing something. I'm curious on seeing your v6.
Thanks
Philipp
>
> > 2. Move the mechanism to run bpf progs from kexec_pe_image.c to
> > kexec_file.c (with the new hook in the file_ops there shouldn't be
> > any problems).
>
> Yes, the logic of loading and runing bpf progs should be placed in
> kexec_file.c.
>
> > 3. Rename CONFIG_KEXEC_PE_IMAGE to CONFIG_KEXEC_BPF
>
> OK.
>
> > 4. Rename kexec_pe_image.c (and the functions within) to
> > kexec_uefi_app.c
> >
>
> OK.
>
> Thanks for your careful review and good suggestion.
>
>
> Best Regards,
>
> Pingfan
>
> > Thanks
> > Philipp
> >
> > On Tue, 19 Aug 2025 09:24:16 +0800
> > Pingfan Liu <piliu at redhat.com> wrote:
> >
> > > Cc systemd-devel at lists.freedesktop.org so any UKI expert can comment
> > >
> > > *** Review the history ***
> > >
> > > Nowadays UEFI PE bootable image is more and more popular on the distribution.
> > > But it is still an open issue to load that kind of image by kexec with IMA enabled
> > >
> > > There are several approaches to reslove this issue, but none of them are
> > > accepted in upstream till now.
> > >
> > > The summary of those approaches:
> > > -1. UEFI service emulator for UEFI stub
> > > -2. PE format parser in kernel
> > >
> > > For the first one, I have tried a purgatory-style emulator [1]. But it
> > > confronts the hardware scaling trouble. For the second one, there are two
> > > choices, one is to implement it inside the kernel, the other is inside the user
> > > space. Both zboot-format [2] and UKI-format [3] parsers are rejected due to
> > > the concern that the variant format parsers will inflate the kernel code. And
> > > finally, we have these kinds of parsers in the user space 'kexec-tools'.
> > >
> > >
> > > *** The approach in this series ***
> > >
> > > This approach allows the various PE boot image to be parsed in the bpf-prog,
> > > as a result, the kexec kernel code to remain relatively stable.
> > >
> > > Benefits
> > > And it abstracts architecture independent part and
> > > the API is limitted
> > >
> > > To protect against malicious attacks on the BPF loader in user space, it
> > > employs BPF lskel to load and execute BPF programs from within the
> > > kernel.
> > >
> > > Each type of PE image contains a dedicated section '.bpf', which stores
> > > the bpf-prog designed to parse the format. This ensures that the PE's
> > > signature also protects the integrity of the '.bpf' section.
> > >
> > >
> > > The parsing process operates as a pipeline. The current BPF program
> > > parser attaches to bpf_handle_pefile() and detaches at the end of the
> > > current stage via disarm_bpf_prog(). The results parsed by the current
> > > BPF program are buffered in the kernel through prepare_nested_pe() and
> > > then delivered to the next stage. For each stage of the pipeline, the
> > > BPF bytecode is stored in the '.bpf' section of the PE file. That means
> > > a vmlinuz.efi embeded in UKI format can be handled.
> > >
> > >
> > > Special thanks to Philipp Rudo, who spent significant time evaluating
> > > the practicality of my solution, and to Viktor Malik, who guided me
> > > toward using BPF light skeleton to prevent malicious attacks from user
> > > space.
> > >
> > > *** Test result ***
> > > Configured with RHEL kernel debug file, which turns on most of locking,
> > > memory debug option, I have not seen any warning or bug for 1000 times.
> > >
> > > Test approach:
> > > -1. compile kernel
> > > -2. get the zboot image with bpf-prog by 'make -C tools/kexec zboot'
> > > -3. compile kexec-tools from https://github.com/pfliu/kexec-tools/pull/new/pe_bpf
> > >
> > > The rest process is the common convention to use kexec.
> > >
> > >
> > > [1]: https://lore.kernel.org/lkml/20240819145417.23367-1-piliu@redhat.com/T/
> > > [2]: https://lore.kernel.org/kexec/20230306030305.15595-1-kernelfans@gmail.com/
> > > [3]: https://lore.kernel.org/lkml/20230911052535.335770-1-kernel@jfarr.cc/
> > > [4]: https://lore.kernel.org/linux-arm-kernel/20230921133703.39042-2-kernelfans@gmail.com/T/
> > >
> > > v4 -> v5
> > > - rebased onto Linux 6.17-rc2
> > > - [1/12], use a separate CONFIG_KEEP_COMPRESSOR to decide the section
> > > of decompressor method
> > > - [10/12], add Catalin's acked-by (Thanks Catalin!)
> > >
> > > v3 -> v4
> > > - Use dynamic allocator in decompression ([4/12])
> > > - Fix issue caused by Identical Code Folding ([5/12])
> > > - Integrate the image generator tool in the kernel tree ([11,12/12])
> > > - Address the issue according to Philipp's comments in v3 reviewing.
> > > Thanks Philipp!
> > >
> > > RFCv2 -> v3
> > > - move the introduced bpf kfuncs to kernel/bpf/* and mark them sleepable
> > > - use listener and publisher model to implement bpf_copy_to_kernel()
> > > - keep each introduced kfunc under the control of memcg
> > >
> > > RFCv1 -> RFCv2
> > > - Use bpf kfunc instead of helper
> > > - Use C source code to generate the light skeleton file
> > >
> > >
> > > *** BLURB HERE ***
> > >
> > > Pingfan Liu (12):
> > > kexec_file: Make kexec_image_load_default global visible
> > > lib/decompress: Keep decompressor when CONFIG_KEEP_COMPRESSOR
> > > bpf: Introduce bpf_copy_to_kernel() to buffer the content from bpf-prog
> > > bpf: Introduce decompressor kfunc
> > > kexec: Introduce kexec_pe_image to parse and load PE file
> > > kexec: Integrate with the introduced bpf kfuncs
> > > kexec: Introduce a bpf-prog lskel to parse PE file
> > > kexec: Factor out routine to find a symbol in ELF
> > > kexec: Integrate bpf light skeleton to load zboot image
> > > arm64/kexec: Add PE image format support
> > > tools/kexec: Introduce a bpf-prog to parse zboot image format
> > > tools/kexec: Add a zboot image building tool
> > >
> > > arch/arm64/Kconfig | 1 +
> > > arch/arm64/include/asm/kexec.h | 1 +
> > > arch/arm64/kernel/machine_kexec_file.c | 3 +
> > > include/linux/bpf.h | 42 ++
> > > include/linux/decompress/mm.h | 7 +
> > > include/linux/kexec.h | 10 +
> > > kernel/Kconfig.kexec | 9 +
> > > kernel/Makefile | 2 +
> > > kernel/bpf/Makefile | 3 +
> > > kernel/bpf/helpers.c | 230 +++++++++
> > > kernel/bpf/helpers_carrier.c | 215 +++++++++
> > > kernel/kexec_bpf/Makefile | 71 +++
> > > kernel/kexec_bpf/kexec_pe_parser_bpf.c | 67 +++
> > > kernel/kexec_bpf/kexec_pe_parser_bpf.lskel.h | 147 ++++++
> > > kernel/kexec_file.c | 88 ++--
> > > kernel/kexec_pe_image.c | 463 +++++++++++++++++++
> > > lib/Kconfig | 3 +
> > > lib/decompress.c | 6 +-
> > > tools/kexec/Makefile | 90 ++++
> > > tools/kexec/pe.h | 177 +++++++
> > > tools/kexec/zboot_image_builder.c | 280 +++++++++++
> > > tools/kexec/zboot_parser_bpf.c | 158 +++++++
> > > 22 files changed, 2029 insertions(+), 44 deletions(-)
> > > create mode 100644 kernel/bpf/helpers_carrier.c
> > > create mode 100644 kernel/kexec_bpf/Makefile
> > > create mode 100644 kernel/kexec_bpf/kexec_pe_parser_bpf.c
> > > create mode 100644 kernel/kexec_bpf/kexec_pe_parser_bpf.lskel.h
> > > create mode 100644 kernel/kexec_pe_image.c
> > > create mode 100644 tools/kexec/Makefile
> > > create mode 100644 tools/kexec/pe.h
> > > create mode 100644 tools/kexec/zboot_image_builder.c
> > > create mode 100644 tools/kexec/zboot_parser_bpf.c
> > >
> > >
> > > base-commit: c17b750b3ad9f45f2b6f7e6f7f4679844244f0b9
> >
>
More information about the kexec
mailing list