Fix docs

[roblabla]

WIP, don't merge yet.

[todo]

Kernel heap memory management

Currently, if the kernel ever OOMs, this will lead to a panic. This is really not ideal. Unfortunately, we depend on a lot of Rust structures that make use of allocation (Arc, Vec, String...). Some of those have methods to customize OOM behavior (Vec and String have try_reserve), but some structures (like Arc) seem to be devoid of any function to avoid OOM.
Maybe we should approach the rust developers, see if they would accept a try_new function on Arc/Rc that would return an AllocErr if it fails.
Alternatively, we could start using our own Arc/Rc forks.

---

https://github.com/roblabla42/KFS/blob/e83bf63b8b517769060ca708e1b7a6f90b264339/kernel/src/heap_allocator.rs#L108-L118

---

This comment was generated by todo based on a TODO comment in e83bf63 in #90. cc @roblabla.

[todo]

gdt::get_main_iopb does not prevent creation of multiple mut ref.

There's currently no guarantee that we don't create multiple &mut pointer to the IOPB region, which would cause undefined behavior. In practice, it should only be used by `i386::process_switch`, and as such, there is never actually two main_iopb active at the same time. Still, it'd be nicer to have safe functions to access the IOPB.

---

https://github.com/roblabla42/KFS/blob/e83bf63b8b517769060ca708e1b7a6f90b264339/kernel/src/i386/gdt.rs#L189-L199

---

This comment was generated by todo based on a TODO comment in e83bf63 in #90. cc @roblabla.

[todo]

Don't panic in Session::reply if active_request is not set.

Session::reply currently asserts that an active session is set. This assertion can be trivially triggered by userspace, by calling the reply_and_receive syscall with reply_target set to a Session that hasn't received any request.

---

https://github.com/roblabla42/KFS/blob/e83bf63b8b517769060ca708e1b7a6f90b264339/kernel/src/ipc/session.rs#L331-L338

---

This comment was generated by todo based on a TODO comment in e83bf63 in #90. cc @roblabla.

[todo]

Don't panic in Session::reply if active_request is not set.

Session::reply currently asserts that an active session is set. This assertion can be trivially triggered by userspace, by calling the reply_and_receive syscall with reply_target set to a Session that hasn't received any request.

---

https://github.com/roblabla42/KFS/blob/7d5bf239fbddadd38f2c8b7abbd70fd9fa785af4/kernel/src/ipc/session.rs#L332-L339

---

This comment was generated by todo based on a TODO comment in 7d5bf23 in #90. cc @roblabla.

[todo]

Don't panic in the oom handler, exit instead.

Panicking may allocate, so calling panic in the OOM handler is a terrible idea.

---

https://github.com/roblabla42/KFS/blob/7d5bf239fbddadd38f2c8b7abbd70fd9fa785af4/libuser/src/lib.rs#L96-L102

---

This comment was generated by todo based on a TODO comment in 7d5bf23 in #90. cc @roblabla.

[todo]

Don't panic in Session::reply if active_request is not set.

Session::reply currently asserts that an active session is set. This assertion can be trivially triggered by userspace, by calling the reply_and_receive syscall with reply_target set to a Session that hasn't received any request.

---

https://github.com/roblabla42/KFS/blob/adee47274ba5c0dc0ef112ce8933becd63d3da77/kernel/src/ipc/session.rs#L332-L339

---

This comment was generated by todo based on a TODO comment in adee472 in #90. cc @roblabla.

[todo]

Don't panic in the oom handler, exit instead.

Panicking may allocate, so calling panic in the OOM handler is a terrible idea.

---

https://github.com/roblabla42/KFS/blob/adee47274ba5c0dc0ef112ce8933becd63d3da77/libuser/src/lib.rs#L96-L102

---

This comment was generated by todo based on a TODO comment in adee472 in #90. cc @roblabla.

[roblabla]

Fixed.