Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Attestation report support with openssl #55

Open
wants to merge 16 commits into
base: main
Choose a base branch
from

Conversation

cclaudio
Copy link
Contributor

@cclaudio cclaudio commented May 11, 2023

Tom suggested to break the PR #35 into two. The first part PR #41 got merged and this is the second part.

The function get_report() can be called to request an attestation report; if the last parameter CertBuf is provided, the certificate chain is stored in the CertBuf.addr provided.

Certificate chain

Make sure the certificate chain is populated in the host, otherwise get_report() will return an empty certificate chain. The sev-guest repository explains how to properly set the certificate chain, however, for testing purpose you can just add some random data to the certificates.

$ git clone https://github.com/AMDESE/sev-guest.git
$ cd sev-guest
$ echo -n "1111" > ark.pem
$ echo -n "2222" > ask.pem
$ echo -n "3333" > vcek.pem
$ make sev-host-set-cert-chain
$ sudo ./sev-host-set-cert-chain -r ark.pem -s ask.pem -v vcek.pem

How to test it

This commit can be applied to test the get_report() function.

If you build the code with make FEATURES=verbose and use the launch-qemu.sh script to launch the guest, you should be able to see messages like this in the console.

INFO: SNP_GUEST_REQUEST msg_type 5 encrypted (96 bytes)
INFO: SNP_GUEST_REQUEST msg_type 6 decrypted (1216 bytes)
INFO: Report, 1184 bytes, vmpl 0
INFO:     report_id: [8b, 82, ed, 43, b9, 0, 50, b6, ec, 26, 38, f7, 33, 73, f7, d0, 1f, 44, 8d, 81, 5e, b6, 22, cb, ad, 7c, f4, 24, 81, d9, dd, c0]
INFO:     report_data: [31, 32, 33, 0, 34, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
INFO: certs sample [63, da, 75, 8d, e6, 64, 45, 64, ad, c5, f4, b9, 3b, e8, ac, cd, 60, 0, 0, 0, 5, 0, 0, 0, 4a, b7, b3, 79, bb, ac, 4f, e4, a0, 2f, 5, ae, f3, 27, c7, 82, 65, 0, 0, 0, 5, 0, 0, 0, c0, b4, 6, a4, a8, 3, 49, 52, 97, 43, 3f, b6, 1, 4c, d0, ae, 6a, 0, 0, 0, 5, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 33, 33, 33, 33, a, 32, 32, 32, 32, a, 31, 31, 31, 31, a, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
INFO: SNP_GUEST_REQUEST msg_type 5 encrypted (96 bytes)
INFO: SNP_GUEST_REQUEST msg_type 6 decrypted (1216 bytes)
INFO: Report, 1184 bytes, vmpl 0
INFO:     report_id: [8b, 82, ed, 43, b9, 0, 50, b6, ec, 26, 38, f7, 33, 73, f7, d0, 1f, 44, 8d, 81, 5e, b6, 22, cb, ad, 7c, f4, 24, 81, d9, dd, c0]
INFO:     report_data: [35, 36, 37, 0, 38, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]

The test basically requests two attestation reports, where the certificate chain is requested only in the first request. The first 600 bytes of the cert chain is printed.

After updating to rustc 1.71, the SVSM build started to fail:

$ rm .prereq
$ make prereq
$ make
...
cryptlib.c:(.text.unlikely+0x8): undefined reference to `abort'
assert.c:(.text+0x22): undefined reference to `abort'
...

The abort function is implemented in Rust and exported for external
libraries. The option "-Wl,-u,abort" prevents it from being removed from
the svsm.elf.

Signed-off-by: Claudio Carvalho <[email protected]>
@cclaudio
Copy link
Contributor Author

One of the github tests is failing because it uses the system glibc to build the code.

@cclaudio
Copy link
Contributor Author

cclaudio commented May 11, 2023

@tlendacky @Zildj1an we don't necessarily need to merge this PR into main. If you create a vtpm feature branch we could just merge this PR into it; @dubek could also submit his PR #44 on top of the branch.

I have another patch set that adds the vtpm changes, I can submit it once this PR is merged.

src/psp/msg_report.rs Outdated Show resolved Hide resolved
Copy link
Contributor

@Zildj1an Zildj1an left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Just some minor tweaks from my side :)

src/psp/guest_request_cmd.rs Outdated Show resolved Hide resolved
src/psp/guest_request_cmd.rs Outdated Show resolved Hide resolved
src/psp/guest_request_cmd.rs Outdated Show resolved Hide resolved
src/psp/guest_request_cmd.rs Outdated Show resolved Hide resolved
src/psp/request.rs Outdated Show resolved Hide resolved
src/psp/msg_report.rs Outdated Show resolved Hide resolved
src/psp/guest_request_cmd.rs Outdated Show resolved Hide resolved
src/psp/guest_request_cmd.rs Outdated Show resolved Hide resolved
src/cpu/vc.rs Outdated Show resolved Hide resolved
src/psp/guest_request_cmd.rs Outdated Show resolved Hide resolved
src/psp/guest_request_cmd.rs Outdated Show resolved Hide resolved
src/psp/guest_request_cmd.rs Outdated Show resolved Hide resolved
src/mem/snpsecrets.rs Outdated Show resolved Hide resolved
src/lib.rs Show resolved Hide resolved
@dubek dubek mentioned this pull request Jun 5, 2023
@cclaudio cclaudio force-pushed the svsm-report-ossl-preview branch 4 times, most recently from e8ec243 to cd01773 Compare June 26, 2023 14:50
@cclaudio
Copy link
Contributor Author

I moved all the openssl calls to a new crate called crypto and marked the crate to be built only when the test feature is not set. With that, all the github tests are passing (including the Tests one).

I will update this PR soon.

There is no need to use the cty package; as of Rust 1.30, the C
primitive types are available in either core::ffi or alloc::ffi.

The cty package and the ffi crate provide similar functionality.

Signed-off-by: Claudio Carvalho <[email protected]>
This removes all the test logic in the wrappers crate and exclude
it entirely from tests.

The wrappers export C functions that already exist in the glibc. It
should not be included in the tests.

Signed-off-by: Claudio Carvalho <[email protected]>
Add bindgen to automatically produce Rust FFI (Foreign Functions
Interface) bindings out of C headers to be used in other crates.

This also renames the target x86_64-unknown-none as bindgen uses it to
properly generate bindings.

Signed-off-by: Claudio Carvalho <[email protected]>
With the attestation report support, the SnpSecrets is now accessed from
multiple places.

Signed-off-by: Claudio Carvalho <[email protected]>
When we request an attestation report to the PSP, the guest crashes with
the following error due to stack overflow.

Launching VM ...
  /tmp/cmdline.866937
char device redirected to /dev/pts/183 (label compat_monitor0)
KVM: entry failed, hardware error 0xffffffff
EAX=00000000 EBX=00000000 ECX=00000000 EDX=00000000
ESI=00000000 EDI=00000000 EBP=00000000 ESP=00000000
EIP=00000000 EFL=00000002 [-------] CPL=0 II=0 A20=1 SMM=0 HLT=0
ES =0000 00000000 00000000 00000000
CS =0000 00000000 00000000 00000000
SS =0000 00000000 00000000 00000000
DS =0000 00000000 00000000 00000000
FS =0000 00000000 00000000 00000000
GS =0000 00000000 00000000 00000000
LDT=0000 00000000 00000000 00000000
TR =0000 00000000 00000000 00000000
GDT=     0000000000000000 00000000
IDT=     0000000000000000 00000000
CR0=80010033 CR2=0000000000000000 CR3=0000000000000000 CR4=00000668
DR0=0000000000000000 DR1=0000000000000000 DR2=0000000000000000 DR3=0000000000000000
DR6=00000000ffff0ff0 DR7=0000000000000400
EFER=0000000000001d00
Code=<??> ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ??
      ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ?? ??

Currently, the SVSM stack size is 4 pages. The same issue happens if we
double the stack size.

This patch increases the BSP and AP stack size to 16 pages.

Signed-off-by: Claudio Carvalho <[email protected]>
Add a generic crypto API required to encrypt and decrypt SNP guest
messages.

This ensures that the nossl.rs stubs are compiled in when we compile the
SVSM with the test feature flag set (e.g. cargo test). This is important
because the test binary is meant to run standalone; so, if we try to
compile any code that calls the SVSM OpenSSL, that will likely cause
compilation conflicts between the libcrt headers and the system glibc.

This also adds a placeholder to when the code is compiled without
setting the test feature flag (e.g. cargo build). Later we can replace
that to point out to a real crypto implementation.

Signed-off-by: Claudio Carvalho <[email protected]>
When we compile the code without setting the test feature flag (e.g.
cargo build), only the OpenSSL-based implementation is compiled in.

Signed-off-by: Claudio Carvalho <[email protected]>
Each SNP guest request command has its own request and response
structure definition. However, they are sent and received from the PSP
wrapped in the SnpGuestRequestMsg structure.

SNP guest request commands are communicated through the hypervisor, so
they are expected to be encrypted with a key known only by the guest and
the PSP.

Signed-off-by: Claudio Carvalho <[email protected]>
This pre-allocates the required SNP_GUEST_REQUEST buffers at boot time
to prevent us from having to deal with out-of-memory issues later at
boot time or runtime.

To send a SNP Guest Request command to the PSP, we need to provide the
request and response physical addresses to the PSP; depending on the
command we also need to provide the physical address of a data buffer.
The size of these buffers are 1, 1 and 4 pages, and they need to be
shared with the hypervisor.

Access to the buffers are controled through a spinlock.

Signed-off-by: Claudio Carvalho <[email protected]>
The function vc_snp_guest_request sends SNP Guest Request messages
following the GHCB protocol defined in the Guest Hypervisor
Communication Block (GHCB) specification. It supports sending either a
extended or non-extended SNP_GUEST_REQUEST message; each one has its own
VMGEXIT exit code.

Signed-off-by: Claudio Carvalho <[email protected]>
As described in the SNP Firmware ABI specification, SNP_GUEST_REQUEST
commands are protected using AES_GCM 256 and a VMPCK key to
encrypt/decrypt the message. The firmware generates and provides the
VMPCK keys to the guest in a especial secrets page that only the guest
and firmware have access to, so the hypervisor cannot alter messages
without detection nor read the plaintext of the message.

We are limiting the send_request() implementation to encrypt/decrypt
messages using the VMPCK0 key only. From guest userspace we can send
SNP_GUEST_REQUEST commands using other VMPCK keys.

The send_request() also supports sending extended SNP_GUEST_REQUEST
commands, where the extended part of it is provided by the hypervisor.
For example, in an extended attestation report the PSP firmware provides
the attestation report and the hypervisor provides the certificate chain
needed to verify the attestation report.

If a SNP_GUEST_REQUEST fails, we disable the VMPCK0 to prevent it from
being used again to send SNP_GUEST_REQUEST commands to the PSP firmware.
The only two exceptions are when the hypervisor returns the error codes
below:

- BUSY: in this case we re-send the message and if it resturns any error
code, we disable the VMPCK0.

- INVALID_DATA_BUFFER_LEN: the data buffer provided in the extended
SNP_GUEST_REQUEST command is too small. So, we re-send the command, but
as non-extended SNP_GUEST_REQUEST in order to protect the IV (sequence
number) from replay attacks. If it returns any error code, we disable
the VMPCK0.

The two cases above could be improved later when, for example, timer
support is added to the SVSM.

Signed-off-by: Claudio Carvalho <[email protected]>
The guest uses the SnpReportRequest structure to assemble an attestation
report request.

The PSP uses the SnpReportResponse structure to assemble an attestation
report response, which contains the signed attestation report requested.

Signed-off-by: Claudio Carvalho <[email protected]>
The get_report() can be used to request an attestation report to the PSP
firmware. If the last parameter CertBuf is provided, the certificate
chain required to verify the attestation report is saved in the
CertBuf.addr provided.

If get_report() fails an Err(error_code) is returned. The wrapped error_code is
compliant with the Core Protocol error codes defined in the SVSM spec 0.62
(draft). The psp_rc reference provided can be used to further understand
the error.

Example:

let buf: x86_64::addr::VirtAddr = mem::mem_allocate(0x4000).unwrap();
let mut certs: CertsBuf = CertsBuf::new(buf, 0x4000usize);
let mut psp_rc: u64 = 0;
let mut data: [u8; USER_DATA_SIZE] = [0u8; USER_DATA_SIZE];
data[0] = 0x31;
data[1] = 0x32;
data[2] = 0x33;
data[4] = 0x34;

// Test extended attestation report request
let result: Result<psp::msg_report::SnpReportResponse, u64> =
    psp::request::get_report(&data, &mut psp_rc, Some(&mut certs));

if let Ok(resp) = result {
    prints!("INFO: Report, {} bytes, vmpl {}\n",
            { resp.get_report_size() },
            { resp.get_report().get_vmpl() }
    );
    prints!("INFO: report_id: {:x?}\n", {
            resp.get_report().get_report_id()
    });
    prints!("INFO: report_data: {:x?}\n", {
            resp.get_report().get_report_data()
    });

    let sample: *const [u8; 500] = buf.as_ptr() as *const [u8; 500];
    prints!("INFO: certs sample {:x?}\n", { unsafe { *sample } });
}

Signed-off-by: Claudio Carvalho <[email protected]>
Add the RUST_INSTALLER_ARGS variable to allow passing parameters to the
Rust installer in the 'make prereq'. For example:

$ make prereq RUST_INSTALLER_ARGS='-y'

The cmdline above tells the rust installer that we want to go with
the default options without any prompt. It could be useful in the github
CI, where tty is not available.

If RUST_INSTALLER_ARGS is not set outside the Makefile, the Rust
installer is executed without any parameter.

All assignments to RUST_INSTALLER_ARGS within the Makefile is ignored,
unless the override directive is used.

Signed-off-by: Claudio Carvalho <[email protected]>
The test.yml calls the same 'cargo test' command line as defined in the
Makefile test target.

This updates the test.yml to call 'make test' so that we don't need to
maintain the same test in multiple places.

Signed-off-by: Claudio Carvalho <[email protected]>
@cclaudio
Copy link
Contributor Author

cclaudio commented Jul 4, 2023

svsm-report-ossl-preview branch updated:

  • Call bindgen in the Makefile rather than having a build script (build.rs). The build.rs does not support checking if we are building for tests. In cargo test, we are running the bindgen generated test cases.
  • Remove cty and bindgen packages from Cargo.toml
  • Move the openssl calls to a separate crate allowing us to compile it out for tests
  • Call make test in the github CI rather than cargo test, this way we don't need to maintain the same test in multiple places.

Updated the test commit link in the PR description.

@tlendacky
Copy link
Collaborator

Hi Claudio, when I try to build your pull request I'm encountering the following error:

   Compiling linux_svsm v0.1.0 (/root/kernels/linux-svsm-claudio-build-x86_64)
error[E0432]: unresolved imports `crate::bindings::EVP_CIPHER_CTX_ctrl`, `crate::bindings::EVP_CIPHER_CTX_free`,
`crate::bindings::EVP_CIPHER_CTX_new`, `crate::bindings::EVP_CIPHER_CTX_set_key_length`,
`crate::bindings::EVP_DecryptFinal_ex`, `crate::bindings::EVP_DecryptInit_ex`, `crate::bindings::EVP_DecryptUpdate`,
`crate::bindings::EVP_EncryptFinal_ex`, `crate::bindings::EVP_EncryptInit_ex`, `crate::bindings::EVP_EncryptUpdate`,
`crate::bindings::EVP_aes_256_gcm`
  --> src/crypto/openssl.rs:15:9
   |
15 |         EVP_CIPHER_CTX_ctrl, EVP_CIPHER_CTX_free, EVP_CIPHER_CTX_new,
   |         ^^^^^^^^^^^^^^^^^^^  ^^^^^^^^^^^^^^^^^^^  ^^^^^^^^^^^^^^^^^^ no `EVP_CIPHER_CTX_new` in `bindings`
   |         |                    |
   |         |                    no `EVP_CIPHER_CTX_free` in `bindings`
   |         no `EVP_CIPHER_CTX_ctrl` in `bindings`
16 |         EVP_CIPHER_CTX_set_key_length, EVP_DecryptFinal_ex, EVP_DecryptInit_ex, EVP_DecryptUpdate,
   |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^  ^^^^^^^^^^^^^^^^^^^  ^^^^^^^^^^^^^^^^^^  ^^^^^^^^^^^^^^^^^ no `EVP_DecryptUpdate` in `bindings`
   |         |                              |                    |
   |         |                              |                    no `EVP_DecryptInit_ex` in `bindings`
   |         |                              no `EVP_DecryptFinal_ex` in `bindings`
   |         no `EVP_CIPHER_CTX_set_key_length` in `bindings`
17 |         EVP_EncryptFinal_ex, EVP_EncryptInit_ex, EVP_EncryptUpdate, EVP_aes_256_gcm,
   |         ^^^^^^^^^^^^^^^^^^^  ^^^^^^^^^^^^^^^^^^  ^^^^^^^^^^^^^^^^^  ^^^^^^^^^^^^^^^
   |         |                    |                   |
   |         |                    |                   no `EVP_EncryptUpdate` in `bindings`
   |         |                    no `EVP_EncryptInit_ex` in `bindings`
   |         no `EVP_EncryptFinal_ex` in `bindings`
   |
help: a similar name exists in the module
   |
15 |         EVP_CIPHER_CTX, EVP_CIPHER_CTX_free, EVP_CIPHER_CTX_new,
   |         ~~~~~~~~~~~~~~
help: a similar name exists in the module
   |
15 |         EVP_CIPHER_CTX_ctrl, EVP_CIPHER_CTX, EVP_CIPHER_CTX_new,
   |                              ~~~~~~~~~~~~~~
help: a similar name exists in the module
   |
15 |         EVP_CIPHER_CTX_ctrl, EVP_CIPHER_CTX_free, EVP_CIPHER_CTX,
   |                                                   ~~~~~~~~~~~~~~

For more information about this error, try `rustc --explain E0432`.
error: could not compile `linux_svsm` (lib) due to previous error
make: *** [Makefile:75: external/libcrt/libcrt.a] Error 101
11:57:08   ERROR: make FEATURES="verbose"

Any idea what I'm doing wrong? Is there another step that I'm missing for building?

@Zildj1an
Copy link
Contributor

Zildj1an commented Aug 1, 2023

I'm also facing that compilation error. Could you please tell me, @cclaudio what version of libssl-dev are you using?

@arkivm
Copy link
Contributor

arkivm commented Aug 6, 2023

I build it on a fresh ubuntu 20.04 image without any issues. Remember to clone the submodules to build the deps.

Here are the versions I have,

openssl, libssl-dev 1.1.1f-1ubuntu2.8
libclang-10-dev 1:10.0.0-4ubuntu1
rustc 1.73.0-nightly

Can you post the complete build log?
Also do you see this file generated as part of the build (external/openssl/libcrypto.a)?

@Zildj1an
Copy link
Contributor

Zildj1an commented Aug 7, 2023

Thanks @arkivm, we had cloned the submodules and I saw external/openssl/libcrypto.a. I suspect a compatibility issue between the versions of OpenSSL, because you're using an older version than recent Ubuntus have (see here) whereas I'm trying to compile using a SUSE distro featuring OpenSSL v3.1.1.

But we should be certain that this is the problem so maybe try replicating from your end.

If that turns out to be the issue, we can discuss the best solution. I don't know how hard would it be to update the code so that it works in more recent versions of OpenSSL. We could require specific versions of OpenSSL, we could use conditional compilation depending on the target (e.g. #[cfg(target_os = "suse")]), we could perhaps do dynamic linking, etc. I don't have strong preferences and would like to hear your opinion since it's your PR.

Here's the log:

make
make -C external/libcrt
make[1]: Entering directory '/machine/linux-svsm/external/libcrt'
make[1]: Nothing to be done for 'all'.
make[1]: Leaving directory '/machine/linux-svsm/external/libcrt'
make -C external/openssl -j$(nproc)
make[1]: Entering directory '/machine/linux-svsm/external/openssl'
make depend && make _all
make[2]: Entering directory '/machine/linux-svsm/external/openssl'
make[2]: Leaving directory '/machine/linux-svsm/external/openssl'
make[2]: Entering directory '/machine/linux-svsm/external/openssl'
make[2]: Leaving directory '/machine/linux-svsm/external/openssl'
make[1]: Leaving directory '/machine/linux-svsm/external/openssl'
bindgen \
	include/bindings.h \
	-o bindgen_out.rs \
	--use-core \
	--rustfmt-configuration-file /machine/linux-svsm/.rustfmt.toml \
	--rust-target "nightly" -- \
		-DOPENSSL_RAND_SEED_NONE \
		-I/machine/linux-svsm/external/libcrt/include \
		-I/machine/linux-svsm/external/openssl/include
   Compiling linux_svsm v0.1.0 (/machine/linux-svsm)
error[E0432]: unresolved imports `crate::bindings::EVP_CIPHER_CTX_ctrl`, `crate::bindings::EVP_CIPHER_CTX_free`, `crate::bindings::EVP_CIPHER_CTX_new`, `crate::bindings::EVP_CIPHER_CTX_set_key_length`, `crate::bindings::EVP_DecryptFinal_ex`, `crate::bindings::EVP_DecryptInit_ex`, `crate::bindings::EVP_DecryptUpdate`, `crate::bindings::EVP_EncryptFinal_ex`, `crate::bindings::EVP_EncryptInit_ex`, `crate::bindings::EVP_EncryptUpdate`, `crate::bindings::EVP_aes_256_gcm`
  --> src/crypto/openssl.rs:15:9
   |
15 |         EVP_CIPHER_CTX_ctrl, EVP_CIPHER_CTX_free, EVP_CIPHER_CTX_new,
   |         ^^^^^^^^^^^^^^^^^^^  ^^^^^^^^^^^^^^^^^^^  ^^^^^^^^^^^^^^^^^^ no `EVP_CIPHER_CTX_new` in `bindings`
   |         |                    |
   |         |                    no `EVP_CIPHER_CTX_free` in `bindings`
   |         no `EVP_CIPHER_CTX_ctrl` in `bindings`
16 |         EVP_CIPHER_CTX_set_key_length, EVP_DecryptFinal_ex, EVP_DecryptInit_ex, EVP_DecryptUpdate,
   |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^  ^^^^^^^^^^^^^^^^^^^  ^^^^^^^^^^^^^^^^^^  ^^^^^^^^^^^^^^^^^ no `EVP_DecryptUpdate` in `bindings`
   |         |                              |                    |
   |         |                              |                    no `EVP_DecryptInit_ex` in `bindings`
   |         |                              no `EVP_DecryptFinal_ex` in `bindings`
   |         no `EVP_CIPHER_CTX_set_key_length` in `bindings`
17 |         EVP_EncryptFinal_ex, EVP_EncryptInit_ex, EVP_EncryptUpdate, EVP_aes_256_gcm,
   |         ^^^^^^^^^^^^^^^^^^^  ^^^^^^^^^^^^^^^^^^  ^^^^^^^^^^^^^^^^^  ^^^^^^^^^^^^^^^
   |         |                    |                   |
   |         |                    |                   no `EVP_EncryptUpdate` in `bindings`
   |         |                    no `EVP_EncryptInit_ex` in `bindings`
   |         no `EVP_EncryptFinal_ex` in `bindings`
   |
help: a similar name exists in the module
   |
15 |         EVP_CIPHER_CTX, EVP_CIPHER_CTX_free, EVP_CIPHER_CTX_new,
   |         ~~~~~~~~~~~~~~
help: a similar name exists in the module
   |
15 |         EVP_CIPHER_CTX_ctrl, EVP_CIPHER_CTX, EVP_CIPHER_CTX_new,
   |                              ~~~~~~~~~~~~~~
help: a similar name exists in the module
   |
15 |         EVP_CIPHER_CTX_ctrl, EVP_CIPHER_CTX_free, EVP_CIPHER_CTX,
   |                                                   ~~~~~~~~~~~~~~

For more information about this error, try `rustc --explain E0432`.
error: could not compile `linux_svsm` (lib) due to previous error
make: *** [Makefile:75: external/libcrt/libcrt.a] Error 101

@arkivm
Copy link
Contributor

arkivm commented Aug 11, 2023

I tried to build a version with OpenSSL v3.1.2 by applying this patch. Seem to compile fine on Ubuntu 20.04 (not run tested).

diff --git a/Makefile b/Makefile
index c146791..577ff3e 100644
--- a/Makefile
+++ b/Makefile
@@ -98,7 +98,7 @@ prereq: .prereq

 external/openssl/Makefile:
        git submodule update --init
-       (cd external/openssl && git checkout OpenSSL_1_1_1q  && \
+       (cd external/openssl && git checkout openssl-3.1.2 && \
                ./Configure \
                        --config=../openssl_svsm.conf \
                        SVSM \

Since it refuses to find any of the imports under crate::bindings on src/crypto/openssl.rs at your end, I am not sure if the bindings_out.rs is correctly generated.

Could you check if you can find these symbols on bindgen_out.rs (should be in the root directory)? I checked the generated output with the below snippet and see all the symbols defined.

for i in $(grep -A 5 "bindings" src/crypto/openssl.rs | grep "^\s\+E" | sed 's/\s\+//g' | tr -d '\n' | tr ',' '\n'); do
  grep -nw "pub fn $i" bindgen_out.rs;
done

script output:

13403:    pub fn EVP_CIPHER_CTX_ctrl(
13388:    pub fn EVP_CIPHER_CTX_free(c: *mut EVP_CIPHER_CTX);
13382:    pub fn EVP_CIPHER_CTX_new() -> *mut EVP_CIPHER_CTX;
13391:    pub fn EVP_CIPHER_CTX_set_key_length(
13099:    pub fn EVP_DecryptFinal_ex(
13065:    pub fn EVP_DecryptInit_ex(
13083:    pub fn EVP_DecryptUpdate(
13043:    pub fn EVP_EncryptFinal_ex(
13016:    pub fn EVP_EncryptInit_ex(
13034:    pub fn EVP_EncryptUpdate(
13618:    pub fn EVP_aes_256_gcm() -> *const EVP_CIPHER;

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

5 participants