Disable PCIe ASPM through kernel command line for iq-8275-evk

[krishnachaitanya-linux]

PCIe ASPM is not working with the current boards, so disable ASPM with kernel command line.

[Mani-Sadhasivam]

As discussed offline, both RB4 and RB8 have some link stability issues with ASPM. But this commit just disables ASPM for all boards, which is wrong. This cmdline param should be passed in the relevant machine config file in conf/machine.

[krishnachaitanya-linux]

ok I will raise for conf/machine/iq-8275-evk.conf as this affecting rb4 for now in new PR and close this one.

[krishnachaitanya-linux]

Instead of creating new one, reopened the same PR with new patch.

[Mani-Sadhasivam]

@krishnachaitanya-linux Don't you need the same change for RB8?

[Mani-Sadhasivam]

Also beware that this cmdline param disables all ASPM states, not just L1ss where the issue is currently. So there is power penalty.

[github-actions]

Test run workflow

Test jobs for commit 3da4aca

• Job 126336 on dragonboard-410c - boot test (dragonboard-410c) 20354026225
• Job 126341 on qcs9100-rb8 - boot test (iq-9075-evk) 20354026225
• Job 126333 on qcs9100 - boot test (qcs9100-ride-sx) 20354026225
• Job 126338 on qrb2210-rb1 - boot test (rb1-core-kit) 20354026225
• Job 126337 on dragonboard-410c - boot test (dragonboard-410c) 20354026225
• Job 126345 on qcs9100-rb8 - boot test (iq-9075-evk) 20354026225
• Job 126334 on qcs6490 - boot test (rb3gen2-core-kit) 20354026225
• Job 126343 on qrb2210-rb1 - boot test (rb1-core-kit) 20354026225
• Job 126340 on dragonboard-820c - boot test (dragonboard-820c) 20354026225
• Job 126344 on qcs8300 - boot test (qcs8300-ride-sx) 20354026225
• Job 126339 on dragonboard-820c - boot test (dragonboard-820c) 20354026225
• Job 126346 on qcs9100 - boot test (qcs9100-ride-sx) 20354026225
• Job 126342 on qcs8300 - boot test (qcs8300-ride-sx) 20354026225
• Job 126335 on qcs6490 - boot test (rb3gen2-core-kit) 20354026225

[test-reporting-app]

Test Results

 14 files  ±0   28 suites  +2   33m 45s ⏱️ - 16m 39s
 50 tests  - 2   50 ✅ + 2  0 💤 ±0  0 ❌  - 4 
304 runs  +8  304 ✅ +16  0 💤 ±0  0 ❌  - 8 

Results for commit 3da4aca. ± Comparison against base commit db27458.

This pull request removes 2 tests.

lava ‑ auto-login-action
lava ‑ minimal-boot

[github-actions]

Test run workflow

Test jobs for commit e331608

• Job 126351 on dragonboard-410c - boot test (dragonboard-410c) 20354169856
• Job 126352 on dragonboard-820c - boot test (dragonboard-820c) 20354169856
• Job 126350 on dragonboard-820c - boot test (dragonboard-820c) 20354169856
• Job 126355 on qcs6490 - boot test (rb3gen2-core-kit) 20354169856
• Job 126360 on qcs8300 - boot test (qcs8300-ride-sx) 20354169856
• Job 126347 on qcs9100-rb8 - boot test (iq-9075-evk) 20354169856
• Job 126353 on qcs6490 - boot test (rb3gen2-core-kit) 20354169856
• Job 126358 on dragonboard-410c - boot test (dragonboard-410c) 20354169856
• Job 126348 on qcs9100 - boot test (qcs9100-ride-sx) 20354169856
• Job 126359 on qcs9100 - boot test (qcs9100-ride-sx) 20354169856
• Job 126349 on qcs9100-rb8 - boot test (iq-9075-evk) 20354169856
• Job 126356 on qcs8300 - boot test (qcs8300-ride-sx) 20354169856
• Job 126357 on qrb2210-rb1 - boot test (rb1-core-kit) 20354169856
• Job 126354 on qrb2210-rb1 - boot test (rb1-core-kit) 20354169856

[krishnachaitanya-linux]

@krishnachaitanya-linux Don't you need the same change for RB8?

we are not hitting this issue in RB8 yet for now, may be due to pwrctrl the athk driver is probing after PCIe controller is enabling L1ss. we can raise the patch for RB8 separately. We are hoping to solve this L1ss issue next week.

[krishnachaitanya-linux]

Also beware that this cmdline param disables all ASPM states, not just L1ss where the issue is currently. So there is power penalty.

There is no other way currently to disable L1ss unless a driver change, like I said in the previous reply, we are hoping to close this by next week, then we will come with proper solution, untill then to unblock test teams we can use this.

[vkraleti]

@krishnachaitanya-linux IMHO using the kernel command line for this purpose is not appropriate. command line is for user configurations, not for masking underlying issues. The correct approach here is to root cause and fix the issues that are causing WLAN failures.

If disabling PCIe ASPM is indeed the ultimate fix, please apply it through the defconfig rather than via the kernel command line.

[lumag]

If disabling PCIe ASPM is indeed the ultimate fix, please apply it through the defconfig rather than via the kernel command line.

I think you mean DT / Driver here.

[ricardosalveti]

@krishnachaitanya-linux IMHO using the kernel command line for this purpose is not appropriate. command line is for user configurations, not for masking underlying issues. The correct approach here is to root cause and fix the issues that are causing WLAN failures.

+1, we should work to avoid any custom kernel command line here, as we want the kernel to work as it would work on any other generic distribution (e.g. with systemready we are not supposed to add custom kernel command line arguments).

So NAK from my side.