centos 7.4 nouveau vs. Dell Inspiron 5577 BIOS

[stuartnlevy]

A Dell Inspiron 5577 laptop (with nVidia GTX 1050 Mobile graphics + Intel 8086:591b graphics) worked fairly well with the "nouveau" driver under CentOS 7.3, but not under 7.4. I suspect the Dell ACPI has a bug which the 7.3 environment could ignore but 7.4 can't. Anyway the result is that the nouveau driver tries and fails to power-on the graphics chipset, then locks up the system - I can't save the contents of dmesg.

(Possibly related: under both 7.3 and 7.4, the elrepo "kmod-nvidia" package is installable, but the X server claims it can't find any graphics devices though the hardware type is listed as supported by "nvidia-detect". I'd be even happier to be able to use nvidia's drivers if it's possible.)

I'd like to be able to run nouveau, at least. Here are symptoms under 7.3, where nouveau does work:
(see 7.3-dmesg-514.26.2.txt)

Code: Select all

[    1.288877] ACPI Warning: \_SB_.PCI0.GFX0._DSM: Argument #4 type mismatch - Found [Buffer], ACPI requires [Package] (20130517/nsarguments-95)
[    1.290005] ACPI Warning: \_SB_.PCI0.PEG0.PEGP._DSM: Argument #4 type mismatch - Found [Buffer], ACPI requires [Package] (20130517/nsarguments-95)
[    1.290859] ACPI Warning: \_SB_.PCI0.PEG0.PEGP._DSM: Argument #4 type mismatch - Found [Buffer], ACPI requires [Package] (20130517/nsarguments-95)
[    1.291790] pci 0000:01:00.0: optimus capabilities: enabled, status dynamic power, hda bios codec supported
[    1.292592] VGA switcheroo: detected Optimus DSM method \_SB_.PCI0.PEG0.PEGP handle
[    1.293531] checking generic (b0000000 7e9000) vs hw (c0000000 10000000)
[    1.293531] checking generic (b0000000 7e9000) vs hw (d0000000 2000000)
[    1.293546] nouveau 0000:01:00.0: enabling device (0006 -> 0007)
[    1.294384] scsi host1: ahci
[    1.294486] nouveau 0000:01:00.0: unknown chipset (137000a1)
[    1.294489] nouveau: probe of 0000:01:00.0 failed with error -12

In comparison, under the 7.4 kernels, if nouveau is not blacklisted, there's an error like this (transcribed from a photo of the screen):

Code: Select all

[   18.093483] Adding 3276799k swap on /dev/sda7 ...
[   26.975128] ACPI Error: Method parse/execution failed [\_SB_.PCI0.PGON] (Node ffff88017cd022a0), AE_AML_INFINITE_LOOP (20130517/psparse-536)
[   26.975878] ACPI Error: Method parse/execution failed [\_SB_.PCI0.PEG0.PG00._ON_] (Node ffff88017ccfc690), AE_AML_INFINITE_LOOP (20130517/psparse-536)
[    26.976605] ACPI: Device [PEG0] failed to transition to D0
[    27.078320] ACPI: Cannot set device to a higher-powered state than parent
[    27.079340] nouveau 0000:01:00.0: Refused to change power state, currently in D3
(a couple more dups of the previous two lines, a few milliseconds later)
[     27.165027] nouveau 0000:01:00.0: DRM: resuming kernel object tree...
[     52.089102] NMI watchdog: BUG: soft lockup - CPU#1 stuck for 22s! [kworker/1:2:244]
[     52.089915] Modules linked in: [i]long list of them[/i]
[     52.094666] CPU: 1 PID: 244 Comm: kworker/1:2 Not tainted 3.10.0-693.2.2.el7.x86_64 #1
[     52.095615] Hardware name: Dell Inc. Inspiron 5577/090HMC, BIOS 1.0.0 02/10/2017
[     52.096572] Workqueue: pm pm_runtime_work
[     52.097557] task: ffff880277213f40 ti: ffff880062dc0000 task.ti: ffff880062dc0000
{I won't copy the register dump here}
[     52.114411] Call Trace:
[     52.115427]   .... nv04_timer_read+0x2b/0x70 [nouveau]
[     52.116431]   .... nvkm_timer_read+0xf/0x20 [nouveau]
[     52.116431]  .... nvkm_pmu_reset+0x78/0x180 [nouveau]
[     52.117447]  ..... nvkm_pu_preinit+0x12/0x20 [nouveau]
[     52.119465] ..... nvkm_subdev_preinit+0x41/0x120 [nouveau]
[     52.120458] ..... nvkm_device_init+0x73/0x200 [nouveau]
... etc etc....
                                nvif_client_resume+0x14/0x20
                                nouveau_do_resume+0x4d/0x140
                                pci_pm_runtime_resume+0x7b/0xc0
...

After this there's a series of lockups at ~23-second intervals, for at least several minutes, and the system is unresponsive except when it's printing the watchdog report.

Do I need to get into ACPI debugging to track this down?!

There is a BIOS update available, and I'll try that too.

[stuartnlevy]

For what it's worth, the Dell BIOS update (from 1.1.0 to 1.1.3) didn't affect the 7.4 nouveau ACPI problem. It still can't power on the device using 7.4 nouveau driver, and still locks up the system when it tries to.