Closed ghost closed 5 years ago
current r345127 and later panic on kldload amdgpu with drm-v4.16/drm-v4.16-fbsd13.0 and graphics/drm-current-kmod
https://share.icloud.com/photos/09OpjtyDsf71uh7xh8ps08wpw
I've had to go back to r345089 to have a working amdgpu. I'm using a Ryzen 7 1700 w/ rx480
I've had to go back to r345089
Do you have the exact revision that breaks it?
Some of the updates for linuxkpi 5.0 went in prematurely. These are r345090 -> r345110. Would be nice to know which one cause the breakage if someone could bisect.. I won't be able to spend time on this until Thursday afternoon UTC.
The drm-v4.16-fbsd13.0 branch is meant to be used when all updates are committed which will be soon.
The only think I can find in base linuxkpi that could cause problems is changes to DMAATTR* macros. This would however not be a problem is drm drivers are rebuilt from source against > r345127. Can you confirm that this is done?
r345101 works with drm-v4.16, 102->115 fail to compile the kms-drm module - 127 compiles the module but panics the system on kldload
I'm here: 345103 good 345106 bad
@gldisater That's just because of a double define. IS_ALIGNED I think. You can easily get it to build if you remove the one in drm code or base.
Hmm.. That is weird. It means that an error that was previously ignored is now detected...
Yep, a lot of them are failing. (return value oid = NULL)
--SNIP--
sysfs_create_dir: name: i2c-5, oid 0xfffff80015b64000
sysfs_create_dir: name: i2c-6, oid 0xfffff80015b63f00
sysfs_create_dir: name: i2c-7, oid 0xfffff80015b63e80
[drm] vm size is 64 GB, 2 levels, block size is 10-bit, fragment size is 9-bit
[drm:gmc_v8_0_init_microcode]
drmn0: successfully loaded firmware image with name: amdgpu/polaris10_mc.bin
drmn0: VRAM: 8192M 0x000000F400000000 - 0x000000F5FFFFFFFF (8192M used)
drmn0: GTT: 256M 0x0000000000000000 - 0x000000000FFFFFFF
Successfully added WC MTRR for [0xe0000000-0xefffffff]: 0;
[drm] Detected VRAM RAM=8192M, BAR=256M
[drm] RAM width 256bits GDDR5
sysfs_create_dir: name: memory_accounting, oid 0
sysfs_create_dir: name: kernel, oid 0
sysfs_create_file: name: kernel, oid 0
sysfs_create_file: name: kernel, oid 0
sysfs_create_file: name: kernel, oid 0
sysfs_create_file: name: kernel, oid 0
sysfs_create_file: name: kernel, oid 0
sysfs_create_dir: name: dma32, oid 0
sysfs_create_file: name: dma32, oid 0
sysfs_create_file: name: dma32, oid 0
sysfs_create_file: name: dma32, oid 0
sysfs_create_file: name: dma32, oid 0
sysfs_create_file: name: dma32, oid 0
[TTM] Zone kernel: Available graphics memory: 3614640 kiB
[TTM] Zone dma32: Available graphics memory: 2097152 kiB
[TTM] Initializing pool allocator
sysfs_create_dir: name: pool, oid 0
sysfs_create_file: name: pool, oid 0
sysfs_create_file: name: pool, oid 0
sysfs_create_file: name: pool, oid 0
sysfs_create_dir: name: buffer_objects, oid 0
sysfs_create_file: name: buffer_objects, oid 0
[drm] amdgpu: 8192M of VRAM memory ready
[drm] amdgpu: 5294M of GTT memory ready.
i_size_write unimplemented
[drm] GART: num cpu pages 65536, num gpu pages 65536
[drm] PCIE GART of 256M enabled (table at 0x000000F400040000).
[drm] Supports vblank timestamp caching Rev 2 (21.10.2013).
--SNIP--
Fixed and tested on 13-CURRENT here https://github.com/FreeBSDDesktop/kms-drm/tree/drm-v4.16 Can someone test this branch on a 12-RELEASE system to make sure nothing's broken? If OK, I'll cherry-pick this fix into the ports branch.
drm-v4.16 does not compile on r345383 - unable to find pci_if.h
@gldisater try now.
works on current for me now, i don't have a 12 BE on this system, but I could probably install one later
Works for me on 12-STABLE
Awesome, thank you!
I have the following problem on 12.0-release (I don't use -RELEASE branch, expect pebkac): kldload /boot/modules/amdgpu.ko KLD amdgpu.ko: depends on ttm - not available or version mismatch
ls -l /boot/modules/ttm.ko ls: /boot/modules/ttm.ko: No such file or directory
cp /usr/ports/graphics/drm-fbsd12.0-kmod/work/stage/boot/modules/ttm.ko /boot/modules/ kldload /boot/modules/amdgpu.ko
It works after that
Yes that is expected since the port is not updated yet. I wanted to confirm that it works on 12.0 first. Thanks for locating the problem and testing! I'll get the port updated today.
It's also ok for me on current
Fixed in ports with https://svnweb.freebsd.org/ports?view=revision&revision=496567
thanks!
Hi, It used to work with a kernel from january but it's broken now ;
FreeBSD 13.0-CURRENT #1 r345299
[drm] amdgpu kernel modesetting enabled. drmn0: on vgapci0
vgapci0: child drmn0 requested pci_enable_io
vgapci0: child drmn0 requested pci_enable_io
[drm] initializing kernel modesetting (OLAND 0x1002:0x6608 0x1028:0x2120 0x00).
[drm] register mmio base: 0xF7E00000
[drm] register mmio size: 262144
[drm] PCI I/O BAR is not found.
[drm] BIOS signature incorrect 0 0
ATOM BIOS: 113-C5790300-101
[drm] vm size is 64 GB, 2 levels, block size is 10-bit, fragment size is 9-bit
drmn0: successfully loaded firmware image with name: radeon/oland_mc.bin
drmn0: VRAM: 2048M 0x000000F400000000 - 0x000000F47FFFFFFF (2048M used)
drmn0: GTT: 1024M 0x0000000000000000 - 0x000000003FFFFFFF
Failed to add WC MTRR for [0xe0000000-0xefffffff]: -22; performance may suffer
[drm] Detected VRAM RAM=2048M, BAR=256M
[drm] RAM width 128bits DDR3
[drm:amdgpu_ttm_global_init] Failed setting up TTM memory accounting subsystem.
[drm:amdgpu_device_ip_init] sw_init of IP block failed -12
drmn0: amdgpu_device_ip_init failed
drmn0: Fatal error during GPU init
[drm] amdgpu: finishing device.
vgapci0: child drmn0 requested pci_disable_io
vgapci0: child drmn0 requested pci_disable_io
device_attach: drmn0 attach returned 12
vgapci0@pci0:1:0:0: class=0x030000 card=0x21201028 chip=0x66081002 rev=0x00 hdr=0x00 vendor = 'Advanced Micro Devices, Inc. [AMD/ATI]' device = 'Oland GL [FirePro W2100]' class = display subclass = VGA