Fix some nitpicks and letfovers from other PRs

[mkow]

Description of the changes

See the commits.

How to test this PR?

CI.

---

This change is 

[woju]

Reviewed 6 of 6 files at r1, all commit messages.
Reviewable status: all files reviewed, 2 unresolved discussions, not enough approvals from maintainers (2 more required), not enough approvals from different teams (1 more required, approved so far: ITL) (waiting on @mkow)

---

pal/src/host/linux-sgx/host_framework.c line 19 at r1 (raw file):

        /* DCAP and upstreamed version used different paths in the past. */
        "/dev/sgx_enclave",
        "/dev/sgx/enclave",

This one probably shouldn't go, it's the one from original DCAP driver, but still compatible, and there's no reason to remove it. Unless we want to instruct people to add custom udev rule if they use old DCAP driver, that will SYMLINK+= it.

---

python/gramine-sgx-sigstruct-view line 52 at r1 (raw file):

        json.dump(sig_readable, output_txt, indent=4)
    else:
        # plaintext format imitates the legacy output of `gramine-sgx-get-token` tool

We've discussed it and resolved that it should remain, because it's still correct that this imitates this (now non-existent) tool: https://reviewable.io/reviews/gramineproject/gramine/2061#-OC6W8vl-J2dK3hV7s4S

[chiache]

Reviewed 5 of 6 files at r1, all commit messages.
Reviewable status: all files reviewed, 3 unresolved discussions, not enough approvals from maintainers (1 more required), not enough approvals from different teams (1 more required, approved so far: ITL) (waiting on @mkow and @woju)

---

pal/src/host/linux-sgx/host_framework.c line 19 at r1 (raw file):

Previously, woju (Wojtek Porczyk) wrote…

This one probably shouldn't go, it's the one from original DCAP driver, but still compatible, and there's no reason to remove it. Unless we want to instruct people to add custom udev rule if they use old DCAP driver, that will SYMLINK+= it.

To me, it's okay to remove this logic if we instruct people to use the new DCAP driver. Potentially, we can add a warning if "/dev/sgx_enclave" does not exist in the system, but "/dev/sgx/enclave" does.

---

pal/src/host/linux-sgx/host_framework.c line 35 at r1 (raw file):

    }
    log_error("Cannot open SGX driver device. Please make sure you're using an up-to-date kernel "
              "or the standalone Intel SGX kernel module is loaded.");

Why not keep this warning in case "/dev/sgx_enclave" does not exist? (i.e., ret = -ENOENT)? This warning is more informative and provide instructions to remedy the issue.

[mkow]

Reviewable status: all files reviewed, 3 unresolved discussions, not enough approvals from maintainers (1 more required), not enough approvals from different teams (1 more required, approved so far: ITL) (waiting on @chiache and @woju)

---

pal/src/host/linux-sgx/host_framework.c line 19 at r1 (raw file):

Previously, chiache (Chia-Che Tsai) wrote…

To me, it's okay to remove this logic if we instruct people to use the new DCAP driver. Potentially, we can add a warning if "/dev/sgx_enclave" does not exist in the system, but "/dev/sgx/enclave" does.

But I think we don't support kernels without the SGX driver built-in, which is incompatible with the DCAP driver?
Also, dropping the support has one more advantage IMO - we won't accidentally get bug reports for Gramine running on the older driver :)

---

pal/src/host/linux-sgx/host_framework.c line 35 at r1 (raw file):

Previously, chiache (Chia-Che Tsai) wrote…

Why not keep this warning in case "/dev/sgx_enclave" does not exist? (i.e., ret = -ENOENT)? This warning is more informative and provide instructions to remedy the issue.

But the error above has all this info, excluding the deprecated part? There's no mention about updating the kernel, because all the supported distros already have it.

---

python/gramine-sgx-sigstruct-view line 52 at r1 (raw file):

Previously, woju (Wojtek Porczyk) wrote…

We've discussed it and resolved that it should remain, because it's still correct that this imitates this (now non-existent) tool: https://reviewable.io/reviews/gramineproject/gramine/2061#-OC6W8vl-J2dK3hV7s4S

Ah, I missed that discussion. But why would we keep this comment if this tool doesn't exist anymore? It doesn't give the reader any useful information?

[chiache]

Reviewable status: all files reviewed, 2 unresolved discussions (waiting on @woju)

---

pal/src/host/linux-sgx/host_framework.c line 19 at r1 (raw file):

Previously, mkow (Michał Kowalczyk) wrote…

But I think we don't support kernels without the SGX driver built-in, which is incompatible with the DCAP driver?
Also, dropping the support has one more advantage IMO - we won't accidentally get bug reports for Gramine running on the older driver :)

I understand we don't support kernels without the SGX driver, but for usability, I think it's useful to provide warning message or hints to users who are using a wrong or incompatible kernel.

---

pal/src/host/linux-sgx/host_framework.c line 35 at r1 (raw file):

Previously, mkow (Michał Kowalczyk) wrote…

But the error above has all this info, excluding the deprecated part? There's no mention about updating the kernel, because all the supported distros already have it.

I think for backward compatibility reasons, some users may be running on an older kernel that do not have this feature. In that case, they will at least know how to fix it, instead of contacting us in the channel.

Anyhow, I am not blocking this PR, but I think in general we can try to provide more warning or hint to users even if they are not exactly following our instructions.

[mkow]

Reviewable status: 5 of 6 files reviewed, 2 unresolved discussions, not enough approvals from maintainers (1 more required), not enough approvals from different teams (1 more required, approved so far: ITL), "fixup! " found in commit messages' one-liners (waiting on @woju)

---

pal/src/host/linux-sgx/host_framework.c line 19 at r1 (raw file):

Previously, chiache (Chia-Che Tsai) wrote…

I understand we don't support kernels without the SGX driver, but for usability, I think it's useful to provide warning message or hints to users who are using a wrong or incompatible kernel.

Ok, makes sense, I added more information to this error, please check now :)

---

pal/src/host/linux-sgx/host_framework.c line 35 at r1 (raw file):

Previously, chiache (Chia-Che Tsai) wrote…

I think for backward compatibility reasons, some users may be running on an older kernel that do not have this feature. In that case, they will at least know how to fix it, instead of contacting us in the channel.

Anyhow, I am not blocking this PR, but I think in general we can try to provide more warning or hint to users even if they are not exactly following our instructions.

Added, see above.