network: fall back to static configuration

[cagatay-y]

No description provided.

[github-actions]

Benchmark Results

Details

Benchmark	Current: 9cebbaf	Previous: 5127c77	Performance Ratio
startup_benchmark Build Time	89.32 s	89.35 s	1.00 ❗
startup_benchmark File Size	0.79 MB	0.79 MB	1.00 ❗
Startup Time - 1 core	0.79 s (±0.04 s)	0.81 s (±0.03 s)	0.98
Startup Time - 2 cores	0.81 s (±0.03 s)	0.82 s (±0.03 s)	0.99
Startup Time - 4 cores	0.82 s (±0.02 s)	0.83 s (±0.02 s)	0.99
multithreaded_benchmark Build Time	86.33 s	92.62 s	0.93 ❗
multithreaded_benchmark File Size	0.89 MB	0.89 MB	1.00 ❗
Multithreaded Pi Efficiency - 2 Threads	91.12 % (±7.60 %)	90.72 % (±8.59 %)	1.00
Multithreaded Pi Efficiency - 4 Threads	45.69 % (±4.15 %)	44.88 % (±4.39 %)	1.02
Multithreaded Pi Efficiency - 8 Threads	26.54 % (±2.67 %)	25.87 % (±3.25 %)	1.03
micro_benchmarks Build Time	92.54 s	100.57 s	0.92 ❗
micro_benchmarks File Size	0.90 MB	0.90 MB	1.00 ❗
Scheduling time - 1 thread	70.53 ticks (±4.97 ticks)	74.98 ticks (±4.47 ticks)	0.94
Scheduling time - 2 threads	40.22 ticks (±4.84 ticks)	40.19 ticks (±4.30 ticks)	1.00
Micro - Time for syscall (getpid)	3.00 ticks (±0.31 ticks)	2.97 ticks (±0.23 ticks)	1.01
Memcpy speed - (built_in) block size 4096	79799.29 MByte/s (±55269.97 MByte/s)	76264.90 MByte/s (±52720.43 MByte/s)	1.05
Memcpy speed - (built_in) block size 1048576	29996.21 MByte/s (±24445.40 MByte/s)	30292.34 MByte/s (±24927.05 MByte/s)	0.99
Memcpy speed - (built_in) block size 16777216	28340.04 MByte/s (±23372.79 MByte/s)	23336.56 MByte/s (±19347.79 MByte/s)	1.21
Memset speed - (built_in) block size 4096	79852.26 MByte/s (±55304.60 MByte/s)	76939.35 MByte/s (±53180.56 MByte/s)	1.04
Memset speed - (built_in) block size 1048576	30705.01 MByte/s (±24826.06 MByte/s)	31021.82 MByte/s (±25328.96 MByte/s)	0.99
Memset speed - (built_in) block size 16777216	29079.79 MByte/s (±23798.00 MByte/s)	24084.23 MByte/s (±19864.53 MByte/s)	1.21
Memcpy speed - (rust) block size 4096	68089.51 MByte/s (±47597.10 MByte/s)	67864.13 MByte/s (±47549.43 MByte/s)	1.00
Memcpy speed - (rust) block size 1048576	29811.72 MByte/s (±24324.34 MByte/s)	30092.92 MByte/s (±24855.28 MByte/s)	0.99
Memcpy speed - (rust) block size 16777216	28250.15 MByte/s (±23253.84 MByte/s)	23650.21 MByte/s (±19669.17 MByte/s)	1.19
Memset speed - (rust) block size 4096	68747.54 MByte/s (±48081.17 MByte/s)	67668.78 MByte/s (±47427.13 MByte/s)	1.02
Memset speed - (rust) block size 1048576	30551.66 MByte/s (±24751.46 MByte/s)	30783.24 MByte/s (±25249.62 MByte/s)	0.99
Memset speed - (rust) block size 16777216	29004.65 MByte/s (±23693.78 MByte/s)	24398.50 MByte/s (±20179.80 MByte/s)	1.19
alloc_benchmarks Build Time	90.89 s	96.87 s	0.94 ❗
alloc_benchmarks File Size	0.86 MB	0.86 MB	1.00 ❗
Allocations - Allocation success	100.00 %	100.00 %	1
Allocations - Deallocation success	100.00 %	100.00 %	1
Allocations - Pre-fail Allocations	100.00 %	100.00 %	1
Allocations - Average Allocation time	6296.31 Ticks (±67.24 Ticks)	5277.51 Ticks (±516.92 Ticks)	1.19 ❗
Allocations - Average Allocation time (no fail)	6296.31 Ticks (±67.24 Ticks)	5277.51 Ticks (±516.92 Ticks)	1.19 ❗
Allocations - Average Deallocation time	1224.60 Ticks (±73.87 Ticks)	959.09 Ticks (±254.64 Ticks)	1.28 ❗
mutex_benchmark Build Time	91.03 s	91.13 s	1.00 ❗
mutex_benchmark File Size	0.90 MB	0.90 MB	1.00 ❗
Mutex Stress Test Average Time per Iteration - 1 Threads	13.34 ns (±0.82 ns)	13.26 ns (±0.96 ns)	1.01
Mutex Stress Test Average Time per Iteration - 2 Threads	20.76 ns (±13.45 ns)	20.28 ns (±12.66 ns)	1.02

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

[Gelbpunkt]

Suggested change

	if !cfg!(feature = "dhcpv4") || hermit_var!("HERMIT_IP").is_some() {
	if cfg!(not(feature = "dhcpv4")) || hermit_var!("HERMIT_IP").is_some() {

Is there a reason we don't want to fall back to 10.0.5.3 if no DHCP configuration is acquired? I feel like that would simplify the logic a bit, and the way I see it, HERMIT_IP is supposed to overwrite the 10.0.5.3 default, having it also imply enabling the fallback is a bit unintuitive. I wouldn't mind always falling back to the static configuration.

[cagatay-y]

My impression is that it is not good practice to claim an IP address without explicit input from the user. When DHCP is not enabled, it is reasonable to infer that the user wants a static IP and if one is not provided use the default, but if DHCP is enabled we shouldn't claim an address without the user asking for it.

Always falling back to the default would also mean that the user has no way of preventing the kernel from claiming an address. I guess we could make it so that when HERMIT_IP is set to something like 0.0.0.0 we don't set an address, but I think the behaviour in the PR is safer.

For the suggested change, I personally prefer to stay in the normal Rust land as opposed to the configuration specification language as much as possible, but I also don't really mind the alternative.

[mkroening]

Thanks! :)