Signed-off-by: danbugs <[email protected]>

In hyperlight-dev#297, we make the base address of guest memory 0x0 instead of 0x200_000 and
remove statically defined memory regions. This commit updates documentation in
accordance to these changes.

Signed-off-by: danbugs <[email protected]>

…yout + removes MemMgrWrapper

This commit completely removes the file w/ the mem layout. There are still
references to it around, but those will be removed in subsequent commits.

With the upcoming refactors, there is no need for the MemMgrWrapper abstraction. This
commit removes in in favour of directly utilizing a SandboxMemoryManager<S> in subsequent
commits.

Signed-off-by: danbugs <[email protected]>

- removes old HyperlightPEB structures (making it more idiomatic Rust).
- removes input/output structures from guest library, moving it to common as they
will need to be used by external guests.
- creates new HyperlightPEB API w/ capabilities to address memory regions from the guest.

Signed-off-by: danbugs <[email protected]>

…uilder

This commit partially introduces the SandboxBuilder mentioned on hyperlight-dev#249, changing
the way we create uninitialized sandboxes.

- adds SandboxBuilder API, which allows for a more flexible memory setup.
- changes to uninit w/ how it is created.
- deletes memory_region.rs file w/ most of it being superseded by the SandboxBuilder structures.
- simplifies config.rs removing all items related to previous static memory layout.

Signed-off-by: danbugs <[email protected]>

… in common library and use CGM

The host now defines a single undifferentiated memory region—the custom guest memory region (or, CGM)—that
the guest can address as it sees fit. This commit changes the guest library to leverage that and perform
said address using the APIs defined in the common library.

Signed-off-by: danbugs <[email protected]>

- removes static layout from SandboxMemoryManager, adds SandboxMemorySections structure built
w/ the SandboxBuilder.
- updates paging to reflect usage of dynamic memory layout.
- interact with common APIs.
- add APIs to allow reading addressed memory sections in PEB from shared memory.

Signed-off-by: danbugs <[email protected]>

… codebase

- temporarily comment out a lot of code (e.g., host/testing/mod.rs).
- temporarily unuse seccomp stuff.
- start using new mgr API and common API.
- pending removal of writting host detail functions to memory.
- change hv_handler to use new mgr API to get guest_function_dispatch_ptr.
- remove layout/memory_region from mod.rs.
- change RawPtr API (should probably be re-evaluated).

Signed-off-by: danbugs <[email protected]>

Updates drivers to map memory region in accordance to CGM usage.

> Note: only updated KVM fully so far. WHP and inprocess drivers are TODOs.

Signed-off-by: danbugs <[email protected]>

Signed-off-by: danbugs <[email protected]>

…d a DebugPrint OutBAction

- the mechanism for causing VM exits and its abstractions over host functions should be
exposed in the common library so they can be used externally.
- 'PrintOutput' should not be a special host function. Instead, we should provide functionality for
debug prints—i.e., the DebugPrint OutBAction.

Signed-off-by: danbugs <[email protected]>

If there's a mismatch in the guest's usize and the host's usize, our input and output
stacks would break. This commit fixes that by using u64 instead of usize for ptr accesses.

Signed-off-by: danbugs <[email protected]>

Removed an uneeded entrypoint argument for the size of the PEB structure, which was a re-use of the ops argument
in dev that is also unused.

+ cleaned up some TODOs (better error messages for unwraps and uneeded host fxn details section).

Signed-off-by: danbugs <[email protected]>

(1) had to add helper impl functions for the HyperlightPEB struct to get the host addresses for some memory regions.
(2) re-added some logic for the LeakedOutBWrapper and for the InProcessDriver.
(3) added an in process test in the sandbox_builder file.
(4) renamed hyperlight_common::hyperlight_peb mod to hyperlight_common::peb.

Signed-off-by: danbugs <[email protected]>

Needed to make some modications to hyperv_linux for it to conform to the new SandboxMemorySections API. With that,
I also had to add a From impl to sandbox_builder. As an aside, I also modified the signature of get_memory_access_violation
to take a borrow of SandboxMemorySections instead of an owned value, which ended up impacting kvm too.

Signed-off-by: danbugs <[email protected]>

Changed whp driver to use the new SandboxMemorySections API.

Signed-off-by: danbugs <[email protected]>

…shdump feature

Crashdump feature used the old MemoryRegion API. Updated it to use the new SandboxMemorySections API (causing
changes in all drivers) and added a test in sandbox_builder for the feature.

Signed-off-by: danbugs <[email protected]>

- cleaned up the HyperlightPEB API in the common library.
-- added MemoryRegion struct to better group related offsets and sizes.
-- removed pub fields from HyperlightPEB struct making fields accessible only via getters/setters.
-- cleaned up, commented, and re-organized existing fxns for HyperlightPEB struct.
- now we modify the rsp in the host if the guest sets up a new stack region (i.e., essentially dropping
the tmp stack).

Signed-off-by: danbugs <[email protected]>

- the guest error data region was removed and now, instead, we leverage the output data region to communicate guest errors back to the host.
- previously, the host error data region was incorrectly accessed making it essentially unused. This commit removes the region altogether. There is
some merit to the region existing in some way, but this is outside of the scope of hyperlight-dev#297.

Signed-off-by: danbugs <[email protected]>

- updated our hello-world example,
- updated the README,
- changed last remaining use of HostPrint to use print (wrapper for using the DebugPrint OutbAction), and
- added an extra test in sandbox_builder.

Signed-off-by: danbugs <[email protected]>

…dler + re-added seccomp test

Previously set mem_access_handler_fn to be NOOP. Brought back original functionality;
albeit a bit questionable.

Signed-off-by: danbugs <[email protected]>

Tests are already covered in the sandbox_builder. In 75b0076, I also added
a test for Windows in-process mode using the load library.

Signed-off-by: danbugs <[email protected]>

- Changed the ABI for the inprocess outb handler from `extern "C"` to
`extern sysv64`. This was needed because `extern "C"` caused issues on Windows.
Previously, we used `extern "win64"`, which was problematic for guests
intending to use `hyperlight_common` and not supporting the Windows ABI.
- Added a test for running inprocess on Windows w/ load library.

Signed-off-by: danbugs <[email protected]>

Signed-off-by: danbugs <[email protected]>