[i2c,dv] Failing block level DV tests as a result of changes.

[alees24]

Description

This issue tracks the state of block level I2C tests, broken or made less reliable by DD work for prod.

Probably failing/unreliable already:

i2c_target_stress_all_with_rand_reset

PR #21621 at this time is thought to be responsible for breaking the following:

• i2c_host_fifo_reset_fmt
• i2c_host_fifo_watermark
• i2c_target_smoke
• i2c_target_stress_rd

Others:

• i2c_intr_test : cip_base_vseq needs 'run_intr_test' changes. (PR #21804)
• cip_base_vseq also requires a change to 'clear_all_interrupts' which is still pending merge and will cause a number of other vseqs to fail, unless it or the final commit of PR #21621 is present.

[andreaskurth]

i2c_host_fifo_watermark is another victim, I think. It passes before the merge of #21621 but fails after (seed 9596103540527566890202921826073542621372745652001006134958711591620971245237 on VCS T-2022.06-SP2). I added it to the list.

[andreaskurth]

i2c_target_stress_rd as well: it passes before the merge of #21621 but fails after (seed 1123967404029634992796962205584092120176243998109053670291080899048047786851 on VCS T-2022.06-SP2). This test simply extends i2c_target_smoke, which may explain why it fails as well.

Added to the list.

[a-will]

These all seem to be passing now--I fixed up several of these along the way.

We're not at 100% at this point, though:

• i2c_csr_bit_bash overflows FIFOs and triggers assertions inside the newer FIFO RTL
• i2c_host_perf has a bizarre calculation for speed that gets affected by stretching that randomly occurs
• i2c_target_unexp_stop appears to not generate correct stimulus
• i2c_target_perf generates incorrect stimulus, with SDA and SCL switching simultaneously
• i2c_target_stretch I haven't looked at

[hcallahan-lowrisc]

Moving tracked effort from this issue to #23077 (This issue can stay open for now, but should be closed when that issue is closed)

[moidx]

Discussed moving as P1 to M5 due to RTL risk analysis performed during last issue triage.

[hcallahan-lowrisc]

This PR doesn't really accurately represent the state of the I2C regressions anymore. Hence I'm going to close it.