feat(riscv): Support advanced interrupt architecture

[sbutz]

Implements #2614.

Feature flag is added in hermit-os/hermit-rs#1077.

Follow-up tasks:

• For each configured msix vector. The external interrupt with the same number is enabled too. This might incur unnecessary cpu traps. Proposed solution: Implement msi vector allocator and store handlers per imsic.
• Bump riscv crate once new version available.
• Specify qemu cpu profile rva23s64 once qemu reaches version 10.1 on ci

[github-actions]

Benchmark Results

Details

Benchmark	Current: c9e016e	Previous: 2e23902	Performance Ratio
startup_benchmark Build Time	92.20 s	80.34 s	1.15 ❗
startup_benchmark File Size	0.79 MB	0.80 MB	0.99 ❗
Startup Time - 1 core	0.74 s (±0.02 s)	0.75 s (±0.02 s)	0.99
Startup Time - 2 cores	0.76 s (±0.03 s)	0.74 s (±0.02 s)	1.04
Startup Time - 4 cores	0.75 s (±0.02 s)	0.74 s (±0.02 s)	1.01
multithreaded_benchmark Build Time	93.20 s	82.11 s	1.14 ❗
multithreaded_benchmark File Size	0.87 MB	0.86 MB	1.02 ❗
Multithreaded Pi Efficiency - 2 Threads	89.61 % (±8.97 %)	85.89 % (±6.61 %)	1.04
Multithreaded Pi Efficiency - 4 Threads	44.39 % (±3.10 %)	43.43 % (±2.56 %)	1.02
Multithreaded Pi Efficiency - 8 Threads	25.56 % (±1.23 %)	25.76 % (±1.53 %)	0.99
micro_benchmarks Build Time	92.65 s	80.40 s	1.15 ❗
micro_benchmarks File Size	0.88 MB	0.86 MB	1.02 ❗
Scheduling time - 1 thread	69.32 ticks (±3.97 ticks)	62.65 ticks (±4.06 ticks)	1.11 ❗
Scheduling time - 2 threads	37.97 ticks (±4.52 ticks)	34.08 ticks (±4.10 ticks)	1.11
Micro - Time for syscall (getpid)	3.89 ticks (±0.61 ticks)	3.45 ticks (±0.58 ticks)	1.13
Memcpy speed - (built_in) block size 4096	77590.11 MByte/s (±53651.66 MByte/s)	82448.38 MByte/s (±56997.13 MByte/s)	0.94
Memcpy speed - (built_in) block size 1048576	29527.85 MByte/s (±24016.75 MByte/s)	30585.98 MByte/s (±24707.84 MByte/s)	0.97
Memcpy speed - (built_in) block size 16777216	24698.97 MByte/s (±20507.99 MByte/s)	26340.06 MByte/s (±21720.96 MByte/s)	0.94
Memset speed - (built_in) block size 4096	78077.57 MByte/s (±53986.30 MByte/s)	82292.76 MByte/s (±56891.50 MByte/s)	0.95
Memset speed - (built_in) block size 1048576	30245.06 MByte/s (±24425.29 MByte/s)	31323.85 MByte/s (±25145.86 MByte/s)	0.97
Memset speed - (built_in) block size 16777216	25393.62 MByte/s (±20942.38 MByte/s)	27104.68 MByte/s (±22209.94 MByte/s)	0.94
Memcpy speed - (rust) block size 4096	69733.19 MByte/s (±48596.37 MByte/s)	74097.96 MByte/s (±51811.44 MByte/s)	0.94
Memcpy speed - (rust) block size 1048576	29454.31 MByte/s (±24010.79 MByte/s)	30361.60 MByte/s (±24602.37 MByte/s)	0.97
Memcpy speed - (rust) block size 16777216	24572.43 MByte/s (±20359.96 MByte/s)	27625.34 MByte/s (±22806.88 MByte/s)	0.89
Memset speed - (rust) block size 4096	70172.07 MByte/s (±48930.90 MByte/s)	74373.47 MByte/s (±51976.48 MByte/s)	0.94
Memset speed - (rust) block size 1048576	30240.74 MByte/s (±24472.45 MByte/s)	31110.89 MByte/s (±25033.24 MByte/s)	0.97
Memset speed - (rust) block size 16777216	25266.40 MByte/s (±20794.03 MByte/s)	28386.93 MByte/s (±23265.03 MByte/s)	0.89
alloc_benchmarks Build Time	86.71 s	74.76 s	1.16 ❗
alloc_benchmarks File Size	0.87 MB	0.87 MB	1.00 ❗
Allocations - Allocation success	91.38 %	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	6250.68 Ticks (±379.89 Ticks)	5860.58 Ticks (±98.43 Ticks)	1.07
Allocations - Average Allocation time (no fail)	7215.59 Ticks (±395.18 Ticks)	6554.81 Ticks (±92.86 Ticks)	1.10 ❗
Allocations - Average Deallocation time	3070.82 Ticks (±564.94 Ticks)	1805.01 Ticks (±250.35 Ticks)	1.70 ❗
mutex_benchmark Build Time	113.38 s	79.82 s	1.42 ❗
mutex_benchmark File Size	0.88 MB	0.86 MB	1.02 ❗
Mutex Stress Test Average Time per Iteration - 1 Threads	13.02 ns (±0.37 ns)	12.10 ns (±0.41 ns)	1.08 ❗
Mutex Stress Test Average Time per Iteration - 2 Threads	89.14 ns (±3.03 ns)	40.26 ns (±1.68 ns)	2.21 ❗

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

[sbutz]

There's one APLIC or PLIC in the system. Therefore locking the global interrupt controller was reasonable.
However if the system fully supports AIA, there's an IMSIC per core and therefore no need to lock the platform level interrupt controller to claim an interrupt.

[sbutz]

Missing fence before ipi.

When software sends IPIs by writing MSIs to the IMSICs of other harts, programmers should consider also the need to execute a FENCE instruction before each store instruction that writes such an MSI. In the absence of FENCEs, many systems guarantee to preserve the order of a hart’s loads and stores only to/from individual devices, not among multiple devices, and not at all for accesses to main memory. With such a system, it must be remembered that each IMSIC is likely to be considered a separate device among the many. For example, if hart A wants to notify hart B that it has completed a task involving accesses to some I/O device, hart A may need to execute a FENCE before sending an MSI to B’s IMSIC, to ensure that all of A’s accesses to the device have actually completed before the MSI could arrive at B. Similarly, if hart A stores anything to memory that should be visible at hart B, a FENCE is likely needed before a subsequent store sending an MSI to B’s IMSIC.

[sbutz]

Closed in favor of smaller PRs: #2731