feat: implement stack protection

[zyuiop]

Isolate stacks in a separate virtual address space, just before the heap.
This ensures that stack overflows cause a page fault instead of silently corrupting the heap (for heap based stacks) or the page table (for the boot stack).

This has been tested only for x86_64 with UEFI, and will definitely not work on other platforms.

TODO:

• Test/Fix on other platforms
• Ensure we do this for all stacks (maybe it's not the case for IDT stacks after the first one)

[github-actions]

Benchmark Results

Details

Benchmark	Current: b5c5ad2	Previous: 2e23902	Performance Ratio
startup_benchmark Build Time	89.88 s	80.34 s	1.12 ❗
startup_benchmark File Size	0.86 MB	0.80 MB	1.08 ❗
Startup Time - 1 core	0.73 s (±0.02 s)	0.75 s (±0.02 s)	0.98
Startup Time - 2 cores	0.74 s (±0.02 s)	0.74 s (±0.02 s)	1.01
Startup Time - 4 cores	0.75 s (±0.02 s)	0.74 s (±0.02 s)	1.02
multithreaded_benchmark Build Time	92.44 s	82.11 s	1.13 ❗
multithreaded_benchmark File Size	0.90 MB	0.86 MB	1.05 ❗
Multithreaded Pi Efficiency - 2 Threads	88.39 % (±9.65 %)	85.89 % (±6.61 %)	1.03
Multithreaded Pi Efficiency - 4 Threads	44.26 % (±3.47 %)	43.43 % (±2.56 %)	1.02
Multithreaded Pi Efficiency - 8 Threads	25.63 % (±1.65 %)	25.76 % (±1.53 %)	1.00
micro_benchmarks Build Time	97.65 s	80.40 s	1.21 ❗
micro_benchmarks File Size	0.90 MB	0.86 MB	1.05 ❗
Scheduling time - 1 thread	64.16 ticks (±3.20 ticks)	62.65 ticks (±4.06 ticks)	1.02
Scheduling time - 2 threads	34.89 ticks (±3.01 ticks)	34.08 ticks (±4.10 ticks)	1.02
Micro - Time for syscall (getpid)	3.77 ticks (±0.65 ticks)	3.45 ticks (±0.58 ticks)	1.09
Memcpy speed - (built_in) block size 4096	81176.67 MByte/s (±56258.31 MByte/s)	82448.38 MByte/s (±56997.13 MByte/s)	0.98
Memcpy speed - (built_in) block size 1048576	30171.36 MByte/s (±24431.69 MByte/s)	30585.98 MByte/s (±24707.84 MByte/s)	0.99
Memcpy speed - (built_in) block size 16777216	26514.13 MByte/s (±21950.41 MByte/s)	26340.06 MByte/s (±21720.96 MByte/s)	1.01
Memset speed - (built_in) block size 4096	80841.29 MByte/s (±56031.38 MByte/s)	82292.76 MByte/s (±56891.50 MByte/s)	0.98
Memset speed - (built_in) block size 1048576	30904.09 MByte/s (±24856.46 MByte/s)	31323.85 MByte/s (±25145.86 MByte/s)	0.99
Memset speed - (built_in) block size 16777216	27255.30 MByte/s (±22407.85 MByte/s)	27104.68 MByte/s (±22209.94 MByte/s)	1.01
Memcpy speed - (rust) block size 4096	72475.66 MByte/s (±50676.10 MByte/s)	74097.96 MByte/s (±51811.44 MByte/s)	0.98
Memcpy speed - (rust) block size 1048576	30008.05 MByte/s (±24356.26 MByte/s)	30361.60 MByte/s (±24602.37 MByte/s)	0.99
Memcpy speed - (rust) block size 16777216	26929.93 MByte/s (±22250.45 MByte/s)	27625.34 MByte/s (±22806.88 MByte/s)	0.97
Memset speed - (rust) block size 4096	72458.85 MByte/s (±50665.29 MByte/s)	74373.47 MByte/s (±51976.48 MByte/s)	0.97
Memset speed - (rust) block size 1048576	30765.46 MByte/s (±24798.95 MByte/s)	31110.89 MByte/s (±25033.24 MByte/s)	0.99
Memset speed - (rust) block size 16777216	27705.69 MByte/s (±22736.00 MByte/s)	28386.93 MByte/s (±23265.03 MByte/s)	0.98
alloc_benchmarks Build Time	83.49 s	74.76 s	1.12 ❗
alloc_benchmarks File Size	0.94 MB	0.87 MB	1.07 ❗
Allocations - Allocation success	91.37 %	91.31 %	1.00 ❗
Allocations - Deallocation success	100.00 %	100.00 %	1
Allocations - Pre-fail Allocations	61.60 %	61.44 %	1.00 ❗
Allocations - Average Allocation time	3994.58 Ticks (±776.93 Ticks)	5860.58 Ticks (±98.43 Ticks)	0.68 ❗
Allocations - Average Allocation time (no fail)	4929.33 Ticks (±735.98 Ticks)	6554.81 Ticks (±92.86 Ticks)	0.75 ❗
Allocations - Average Deallocation time	1284.58 Ticks (±181.07 Ticks)	1805.01 Ticks (±250.35 Ticks)	0.71 ❗
mutex_benchmark Build Time	143.49 s	79.82 s	1.80 ❗
mutex_benchmark File Size	0.90 MB	0.86 MB	1.05 ❗
Mutex Stress Test Average Time per Iteration - 1 Threads	12.16 ns (±0.42 ns)	12.10 ns (±0.41 ns)	1.00
Mutex Stress Test Average Time per Iteration - 2 Threads	65.34 ns (±3.17 ns)	40.26 ns (±1.68 ns)	1.62 ❗

This comment was automatically generated by workflow using github-action-benchmark.

[zyuiop]

Okay I have spent some time trying to port to arm/riscV

[zyuiop]

I extracted the first two commits to #2466 so they can be merged independently :)

[zyuiop]

Note than on ARM/RISC, the boot stack is not moved as we do in x86_64, so it is not protected.

I have a branch for RISCV on my PC, I can push it if there is interest.

[zyuiop]

I can reproduce the Laplace iterations problem locally on main too, so I don't think this is my fault

[mkroening]

Could you rebase and squash any fixup commits? I believe some are already merged, though.

[zyuiop]

my local rustfmt disagrees with the CI, how??

[zyuiop]

I had to reduce the task stack size, otherwise the multi-processor/multi-task uhyve task was failing (4MB available for stacks was not enough when each stack takes 1MB).

In practice, user tasks spawned by the hermit-rs still use the old value, so this fix just frees-up the necessary 512KB for the multiple guard pages.

[stlankes]

@zyuiop Can you resolve the conflicts.