Adam Madsen
841f8a507b
[core] bump dkms version.
2020-11-18 17:56:11 -06:00
Adam Madsen
c5cb424fec
[amd] polaris10: check if reset is needed, use BACO if possible
2020-11-18 17:45:51 -06:00
Adam Madsen
838ccab480
[amd] firmware: fix IOREG access
2020-11-18 17:44:50 -06:00
Adam Madsen
ed604594d3
[amd] common: fix indirect register access
2020-11-18 17:43:28 -06:00
Adam Madsen
864b808696
[core] since hook is the default, add NOHOOK argument to Make
2020-11-18 17:42:46 -06:00
Adam Madsen
7d18e6a4c2
[amd] navi10: actually include that missing header file
2020-11-14 18:39:11 -06:00
Adam Madsen
8818587b84
[amd] navi10: add missing header file.
2020-11-14 18:31:22 -06:00
Adam Madsen
47022fe550
[amd] navi10: a little tidying up
2020-11-14 11:16:50 -06:00
Adam Madsen
0dab58276f
[amd] common: add missing device dereference.
2020-11-14 01:07:52 -06:00
Adam Madsen
246f57e0b0
[core] hook: prefer included fixes over pci/quirks.c
2020-11-13 22:57:16 -06:00
Adam Madsen
8d6eac1f49
[core] fix typo in readme
2020-11-13 22:52:54 -06:00
Adam Madsen
8d2b2f3d4d
[core] Bump DKMS version.
2020-11-13 19:42:48 -06:00
Adam Madsen
331cabae19
[amd] vega10: do not disable some features that apparently cause the SMU to hang up in some circumstances.
2020-11-13 19:42:14 -06:00
Adam Madsen
d99cee4489
[core] bump DKMS module version
2020-11-13 17:36:28 -06:00
Adam Madsen
f20d8b70bc
[amd] vega10: SMU feature disable + gfx reset + BACO seems to be the secret sauce
2020-11-13 17:35:03 -06:00
Adam Madsen
c18730a60d
[amd] common: Map port I/O space (proper way to access certain ATOM registers?)
2020-11-13 17:35:03 -06:00
Adam Madsen
d537ed67d9
[amd] common: remove mutexes and locks
2020-11-13 17:35:03 -06:00
Adam Madsen
edda7f02a5
[core] hooks: zero stack-allocated struct.
2020-11-11 14:04:17 -06:00
Adam Madsen
83ca5e9ae5
[core] Zero stack-allocated struct.
2020-11-11 14:03:30 -06:00
Adam Madsen
6dec3dcbaa
[core] Add device to Navi14 list
2020-11-11 13:13:17 -06:00
Adam Madsen
986e81c09e
[amd] navi10: Fix PSP bootloader read at reset check
2020-11-11 13:12:52 -06:00
Adam Madsen
7b2f5d76dd
Small QoL logging clarification
2020-11-10 13:01:50 -06:00
Adam Madsen
67422cc5bd
Log when hooks are installed, set install_hook=y by default
2020-11-10 13:00:04 -06:00
Adam Madsen
346e99246b
Fix conditional guard against amdgpu module
2020-11-10 11:36:29 -06:00
Adam Madsen
1394ebdd37
Do not reset when device is under control of amdgpu.
2020-11-10 11:10:48 -06:00
Adam Madsen
8d5922b434
AMD: I think that's Navi.
2020-11-10 09:48:36 -06:00
Adam Madsen
1e5120caa4
Add polaris 11 and polaris 12 device IDs
2020-11-10 07:18:48 -06:00
Adam Madsen
9395482980
Remove unneeded logs
2020-11-10 07:15:09 -06:00
Adam Madsen
f3cbc30bf0
Add DKMS config (thanks gnif)
2020-11-10 07:08:05 -06:00
Adam Madsen
5615523f26
Add install target to install module
2020-11-10 06:59:40 -06:00
Adam Madsen
25827381f4
Change exit order in attempt to prevent misc device kobj from sticking around
2020-11-10 06:59:30 -06:00
Adam Madsen
c21bcddd9e
AMD: disable bus resets.
2020-11-10 00:20:32 -06:00
Adam Madsen
2165913cef
Add option to install hook via make load HOOK=1
2020-11-10 00:11:26 -06:00
Adam Madsen
2649176b4d
userspace: Use always-y with userprogs to ensure it builds on 5.8.
2020-11-10 00:09:16 -06:00
Adam Madsen
81516a3241
Add additional source
2020-11-09 22:35:10 -06:00
Adam Madsen
74ea6dd018
Add option to hook device reset through ftrace
2020-11-09 22:33:15 -06:00
Adam Madsen
afbc509011
AMD: allocate amd_fake_dev struct for polaris10.
2020-11-09 14:24:11 -06:00
Adam Madsen
2d08232dc2
AMD: apply polaris ops to polaris cards :)
2020-11-09 14:13:30 -06:00
Adam Madsen
c46c1dd071
AMD: Add PCIE reset method for Polaris 10.
2020-11-09 13:26:16 -06:00
Adam Madsen
91ad547e08
Parse PCI devices in DBDF form.
2020-11-03 18:00:40 -06:00
Adam Madsen
d0b17b2bb5
[amd] Update vega10 to use fake dev init
2020-11-02 15:23:53 -06:00
Adam Madsen
710b434e85
Remove extra tab
2020-11-02 15:18:43 -06:00
Adam Madsen
68ca4816b6
Load and parse BIOS/firmware where possible
2020-11-02 14:47:13 -06:00
Adam Madsen
7faae1b95a
Guard against common cases where vega10 does not
...
need to be reset.
Specifically, check if the card is already in BACO. If so, exit. If the card shows no signs of life, exit because we likely can't do anything. If the card shows signs of life, do a BACO reset.
2020-11-01 23:29:53 -06:00
Adam Madsen
eb42a21118
Vega10 BACO reset should work now.
2020-11-01 21:41:46 -06:00
Adam Madsen
4ecfddfc0e
Unmap mmio post reset.
2020-11-01 21:16:22 -06:00
Adam Madsen
95bcd0f0f1
Collect some precondition info
2020-11-01 20:18:39 -06:00
Adam Madsen
b7554f9035
Free mutex
2020-11-01 19:17:30 -06:00
Adam Madsen
db50c45f35
Store pointer to vendor struct, fix crash in lock.
2020-11-01 19:16:25 -06:00
Adam Madsen
b8517880ea
Implement BACO for vega10.
...
Additionally, imported a whole bunch of stuff from the `amdgpu` module.
2020-11-01 18:48:10 -06:00
Adam Madsen
1a32bb50ba
Add missing license header.
2020-11-01 16:15:29 -06:00
Adam Madsen
9c2185f113
Add structure for common pre- and post-reset tasks
2020-11-01 16:14:11 -06:00
Adam Madsen