catalyst vs open source

[agd5f]

kernel 3.16 was released a year ago. It didn't have the id. It was added in May as bridgman pointed out. You'll need a recent ddx, mesa, libdrm, and kernel. The kernel patch went to stable as well so it should show up in any currently maintained stable trees. AFAIK, 3.16 is not one of them however.

[ossuser]

The ID was added in May so you'll need a fairly recent X driver:

http://cgit.freedesktop.org/xorg/dri...f7583859cca63d

Thanks, bridgman.

Would copying from for example R7 260 in the current debian driver work for the R7 360 ?
Or is there too much difference from 260 to 360 (or their firmware) ?

Just trying to stay on Debian Stable here ;-)

[ossuser]

kernel 3.16 was released a year ago. It didn't have the id. It was added in May as bridgman pointed out. You'll need a recent ddx, mesa, libdrm, and kernel. The kernel patch went to stable as well so it should show up in any currently maintained stable trees. AFAIK, 3.16 is not one of them however.

Thanks, agd5f,

I could check first for any backported kernel/X/mesa etc, when available.
That might save me quite some time.

But if only the ID (and not a lot of code) was added, then patching should be ok.
I guess I have some digging to do...

[ossuser]

Ok, so I found in include/drm/drm_pciids.h (linux-4.2-rc1 kernel sources):

[SIZE=14px]{0x1002, 0x665f, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_BONAIRE|RADEON_NEW_MEMMAP}, \[/SIZE]

... which I can add in my current kernel sources.

Funny enough, this line is also found (without bitwise orred RADEON_NEW_MEMMAP) in drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c

Hmmm, R7 360 in AMDGPU DRM ? I wouldn't complain about that.

[bridgman]

Funny enough, this line is also found (without bitwise orred RADEON_NEW_MEMMAP) in drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c

Hmmm, R7 360 in AMDGPU DRM ? I wouldn't complain about that.

The amdgpu driver was developed on CI hardware (mostly Kaveri & Bonaire IIRC) and we didn't strip out the CI code, but it's disabled by default in upstream & distro builds because we already had upstream support in radeon. Production use will be Tonga, Carizzo, Fiji and up.

[ossuser]

Yeah, I kinda expected that answer.

Sorry I made you repeat the message once more.

[ossuser]

Patching the R7 360 PCI-ID in the drm_pciids.h header file seems to work.

 # dmesg | grep adeon
[    3.831305] [drm] radeon kernel modesetting enabled.
[    3.833508] radeon 0000:02:00.0: [COLOR=#FF0000]Invalid ROM contents[/COLOR]
[    3.833607] radeon 0000:02:00.0: VRAM: 2048M 0x0000000000000000 - 0x000000007FFFFFFF (2048M used)
[    3.833608] radeon 0000:02:00.0: GTT: 1024M 0x0000000080000000 - 0x00000000BFFFFFFF
[    3.833670] [drm] radeon: 2048M of VRAM memory ready
[    3.833670] [drm] radeon: 1024M of GTT memory ready.
[    3.838687] radeon 0000:02:00.0: firmware: direct-loading firmware radeon/BONAIRE_pfp.bin
[    3.839326] radeon 0000:02:00.0: firmware: direct-loading firmware radeon/BONAIRE_me.bin
[    3.841506] radeon 0000:02:00.0: firmware: direct-loading firmware radeon/BONAIRE_ce.bin
[    3.843087] radeon 0000:02:00.0: firmware: direct-loading firmware radeon/BONAIRE_mec.bin
[    3.843541] radeon 0000:02:00.0: firmware: direct-loading firmware radeon/BONAIRE_rlc.bin
[    3.844437] radeon 0000:02:00.0: firmware: direct-loading firmware radeon/BONAIRE_sdma.bin
[    3.845047] radeon 0000:02:00.0: firmware: direct-loading firmware radeon/BONAIRE_mc2.bin
[    3.845121] [drm] radeon/BONAIRE_mc2.bin: 31792 bytes
[    3.846603] radeon 0000:02:00.0: firmware: direct-loading firmware radeon/BONAIRE_smc.bin
[    3.857310] [drm] radeon: dpm initialized
[    3.858877] radeon 0000:02:00.0: firmware: direct-loading firmware radeon/BONAIRE_uvd.bin
[    3.860321] radeon 0000:02:00.0: firmware: direct-loading firmware radeon/BONAIRE_vce.bin
[    3.875815] radeon 0000:02:00.0: WB enabled
[    3.875880] radeon 0000:02:00.0: fence driver on ring 0 use gpu addr 0x0000000080000c00 and cpu addr 0xffff8801b6aadc00
[    3.875947] radeon 0000:02:00.0: fence driver on ring 1 use gpu addr 0x0000000080000c04 and cpu addr 0xffff8801b6aadc04
[    3.876013] radeon 0000:02:00.0: fence driver on ring 2 use gpu addr 0x0000000080000c08 and cpu addr 0xffff8801b6aadc08
[    3.876079] radeon 0000:02:00.0: fence driver on ring 3 use gpu addr 0x0000000080000c0c and cpu addr 0xffff8801b6aadc0c
[    3.876145] radeon 0000:02:00.0: fence driver on ring 4 use gpu addr 0x0000000080000c10 and cpu addr 0xffff8801b6aadc10
[    3.877045] radeon 0000:02:00.0: fence driver on ring 5 use gpu addr 0x0000000000076c98 and cpu addr 0xffffc900112b6c98
[    3.877209] radeon 0000:02:00.0: fence driver on ring 6 use gpu addr 0x0000000080000c18 and cpu addr 0xffff8801b6aadc18
[    3.878375] radeon 0000:02:00.0: fence driver on ring 7 use gpu addr 0x0000000080000c1c and cpu addr 0xffff8801b6aadc1c
[    3.878558] radeon 0000:02:00.0: irq 72 for MSI/MSI-X
[    3.878578] radeon 0000:02:00.0: radeon: using MSI.
[    3.878670] [drm] radeon: irq initialized.
[    4.112592] [drm] Radeon Display Connectors
[    4.162533] fbcon: radeondrmfb (fb0) is primary device
[    4.222371] radeon 0000:02:00.0: fb0: radeondrmfb frame buffer device
[    4.222372] radeon 0000:02:00.0: registered panic notifier
[    4.237670] [drm] Initialized radeon 2.39.0 20080528 for 0000:02:00.0 on minor 0

I'm writing this from a radeon driven desktop on Debian Jessie, kernel 3.16

Maybe it needs a little bit more work, hence the ' Invalid ROM contents' message.

The X.log seems a little strange also:

[     4.391] (II) xfree86: Adding drm device (/dev/dri/card0)
[     4.393] (--) PCI:*(0:2:0:0) 1002:665f:174b:e258 rev 129, Mem @ 0xd0000000/268435456, 0xcf800000/8388608, 0xfbc80000/262144, I/O @ 0x0000c000/256, BIOS @ [COLOR=#FF0000]0x????????[/COLOR]/131072

Any idea for improvements, bridgman or agd5f ?

[bridgman]

Looks like you can ignore the "Invalid ROM contents" message as long as the driver loads OK. AFAICS the driver tries a few different mechanisms to access the VBIOS ROM and each failed attempt generates one of these messages.

http://www.phoronix.com/forums/forum/linux-graphics-x-org-drivers/open-source-amd-linux/45819-radeon-0000-01-00-0-invalid-rom-contents?p=580599#post580599

[ossuser]

Ok, for all Linux users who bought Rx 3xx rebranded cards, a final question, bridgman.

As it seems to work well, would it be safe to 'backport' the new PCI-ID 's (+struct info) to older kernels drm_pciids.h (at least 3.16) ?
I'd like to know your and Alex's opinion on this.

I quess you saw this one coming ;-)

[bridgman]

Yep, and Alex already answered it, see post 61 above

The change has (in theory at least) been tagged for stable releases so will get pulled automatically into stable kernels... it's just that 3.16 isn't one of the kernels picked for stable updates AFAIK.

That said, when I looked in the latest stable kernels I didn't actually see it, so we might want to check that it didn't get missed somehow.