Any success stories with Kaveri?

andy713 · 02-04-2014, 10:37 AM

Because I apparently watch too much Top Gear, I thought "how hard could it be?" and "what could possibly go wrong?" and built me a new Kaveri box this week. This was current three weeks ago (with multilib), I upped the kernel version to 3.13.1 and the ATI Catalyst driver to the 14.1 beta. I've been using nvidia the last few years so I'm new to ATI. A few questions:
1. What are the sensor modules? sensors-detect didn't pick up anything.
2. Has /usr/lib64/fglrx/switchlibGL always been broken for Slackware?
3. I did get DRI working in spite of the libGL brokeness, but now google earth 7 crashes right out of the gate with signal 11. GE 6 doesn't have this problem, but it sure is slow.
4. Something in my system doesn't like b43 hardware. I got this during boot:
[ 19.100324] b43-phy0: Loading firmware version 666.2 (2011-02-23 01:15:07)
[ 19.173866] AMD-Vi: Event logged [IO_PAGE_FAULT device=00:14.4 domain=0x0014 address=0x0000000000117040 flags=0x0050]
It then proceeded to spit out many IO_PAGE_FAULT messages per second until rebooted. I had to swap out the wireless card with an Atheros I had lying around to make it stop.

Andy

granth · 02-04-2014, 12:25 PM

Quote:

Originally Posted by andy713

What are the sensor modules? sensors-detect didn't pick up anything.

New board; probably needs new drivers for those sensors.

Quote:

Originally Posted by andy713

Something in my system doesn't like b43 hardware. I got this during boot:
[ 19.100324] b43-phy0: Loading firmware version 666.2 (2011-02-23 01:15:07)
[ 19.173866] AMD-Vi: Event logged [IO_PAGE_FAULT device=00:14.4 domain=0x0014 address=0x0000000000117040 flags=0x0050]
It then proceeded to spit out many IO_PAGE_FAULT messages per second until rebooted. I had to swap out the wireless card with an Atheros I had lying around to make it stop.

Looks like something does not play nice with the IOMMU. You could try turning that off in the bios/efi.

andy713 · 02-07-2014, 09:13 PM

Well, maybe 3.13.2 will fix the CPU temp issue:

hwmon: (k10temp) Add support for Kaveri CPUs
commit d303b1b5fbb688282bbf72a534b9dfed7af9fe4f upstream.

I'll know shortly.

32-bit Google Earth 7 works. Apparently the so-called 64-bit version is still a work in progress.

As for the AMD-Vi IO_PAGE_FAULT issue, I tried all sorts of BIOS settings to make it go away. No effect. There are some fixes for b43 in the 3.13.2 kernel, but I won't be testing those.

JamesGT · 02-07-2014, 09:15 PM

Interesting...my next system will be a Kaveri system. I am currently using a Llano system and it's been great.

andy713 · 02-09-2014, 10:17 AM

So, the 3.13.2 kernel does have Kaveri temp support. Just one problem:

$ sensors
k10temp-pci-00c3
Adapter: PCI adapter
temp1: +1.6°C (high = +70.0°C)
(crit = +70.0°C, hyst = +69.0°C)

Is a wrong number is better than none at all?

granth · 02-10-2014, 10:59 AM

I get similar results with my Vishera CPU using kernel 3.10.17. The values jump around in sync with the power readings. They always are close to half of the power values.

Code:

$ sensors
k10temp-pci-00c3
Adapter: PCI adapter
temp1:        +17.1�C  (high = +70.0�C)
                       (crit = +90.0�C, hyst = +87.0�C)

fam15h_power-pci-00c4
Adapter: PCI adapter
power1:       30.68 W  (crit =  94.99 W)

Code:

$ sensors
k10temp-pci-00c3
Adapter: PCI adapter
temp1:        +21.0�C  (high = +70.0�C)
                       (crit = +90.0�C, hyst = +87.0�C)

fam15h_power-pci-00c4
Adapter: PCI adapter
power1:       46.24 W  (crit =  94.99 W)

Code:

$ sensors
k10temp-pci-00c3
Adapter: PCI adapter
temp1:        +24.8�C  (high = +70.0�C)
                       (crit = +90.0�C, hyst = +87.0�C)

fam15h_power-pci-00c4
Adapter: PCI adapter
power1:       53.39 W  (crit =  94.99 W)

rvdboom · 02-10-2014, 01:22 PM

I also have weird temperature readings. I guess the kernel support is not 100% bug free yet.
Apart from that, I'm using git versions of libdrm, mesa, mesa-glu, glamor and xf86-video-ati and if works very well with proper compositing in KWin, vdpau support using the UVD chip.
Honestly, the chip has been released a month ago, and almost everything works well. I'm impressed.

andy713 · 02-10-2014, 08:23 PM

With the latest kernel and beta Catalyst driver I get this in dmesg:

[ 29.189877] fglrx_pci 0000:00:01.0: irq 98 for MSI/MSI-X
[ 29.190417] BUG: using smp_processor_id() in preemptible [00000000] code: X/864
[ 29.190491] caller is kas_spin_lock+0x1b/0xf0 [fglrx]
[ 29.190494] CPU: 3 PID: 864 Comm: X Tainted: P O 3.13.2-asj #1
[ 29.190496] Hardware name: MSI MS-7721/A88XM-E45 (MS-7721), BIOS V25.1 12/06/2013
[ 29.190498] 0000000000013440 ffff88042ba0b3c8 ffffffff814f98aa 0000000000000003
[ 29.190500] ffff88042ba0b3e0 ffffffff812cc9bc ffff88042ba0b440 ffff88042ba0b400
[ 29.190503] ffffffffa04e8c9b ffff8804231b87d0 ffff8804231af7d0 ffff88042ba0b488
[ 29.190506] Call Trace:
[ 29.190522] [<ffffffff814f98aa>] dump_stack+0x4e/0x7a
[ 29.190526] [<ffffffff812cc9bc>] debug_smp_processor_id+0xcc/0xe0
[ 29.190545] [<ffffffffa04e8c9b>] kas_spin_lock+0x1b/0xf0 [fglrx]
[ 29.190567] [<ffffffffa04ed3d6>] KAS_InterlockedList_InsertAtTail+0x76/0x160 [fglrx]
[ 29.190595] [<ffffffffa0508ba3>] MCIL_InterlockedList+0x193/0x2a0 [fglrx]
[ 29.190667] [<ffffffffa06217f6>] ? FIFO_Queue_Initialize+0xe6/0x100 [fglrx]
[ 29.190693] [<ffffffffa04ff2a5>] ? fireglPcsCommand+0xa5/0x140 [fglrx]
[ 29.190725] [<ffffffffa0506708>] ? MCIL_GetRegistrykey+0x268/0x750 [fglrx]
[ 29.190785] [<ffffffffa062d462>] ? AvailableCallbacksQueue_Initialize+0x52/0x70 [fglrx]
[ 29.190845] [<ffffffffa06306ba>] ? InitializeExtension+0x25a/0x310 [fglrx]
[ 29.190904] [<ffffffffa0630e78>] ? InitializeInterrupts+0x38/0x60 [fglrx]
[ 29.190970] [<ffffffffa061fcac>] ? IRQMGR_InitializeIRQMGR+0x10c/0x120 [fglrx]
[ 29.191034] [<ffffffffa062af00>] ? SPECTRE_QueryAsicTimestamp+0xa0/0xa0 [fglrx]
[ 29.191092] [<ffffffffa062af20>] ? SPECTRE_GetMicrosecRefFreq+0x20/0x20 [fglrx]
[ 29.191150] [<ffffffffa062af40>] ? SPECTRE_IsTimerRunning+0x20/0x20 [fglrx]
[ 29.191208] [<ffffffffa062ae60>] ? SPECTRE_EnableDisplayPortSinkStatusUpdate+0x90/0x90 [fglrx]
[ 29.191266] [<ffffffffa062ad50>] ? SPECTRE_GetHotPlugConnectivityState+0x90/0x90 [fglrx]
[ 29.191325] [<ffffffffa062aff0>] ? SPECTRE_AckTimerInterrupt+0x10/0x10 [fglrx]
[ 29.191389] [<ffffffffa062a850>] ? SPECTRE_Enable_HotPlug+0x190/0x190 [fglrx]
[ 29.191457] [<ffffffffa062acc0>] ? SPECTRE_SetHotPlugConnectivityState+0x470/0x470 [fglrx]
[ 29.191533] [<ffffffffa0635fc0>] ? TAHITI_IVRing_UpdateReadPointer+0x80/0x80 [fglrx]
[ 29.191601] [<ffffffffa0636290>] ? BONAIRE_IVRing_GetPointers+0xc0/0xc0 [fglrx]
[ 29.191659] [<ffffffffa06361d0>] ? BONAIRE_IVRing_InitHardware+0x210/0x210 [fglrx]
[ 29.191717] [<ffffffffa0636340>] ? BONAIRE_IVRing_EnableInterrupts+0xb0/0xb0 [fglrx]
[ 29.191777] [<ffffffffa062b1c0>] ? SPECTRE_Enable_VCE_TRAP_Int+0x80/0x80 [fglrx]
[ 29.191846] [<ffffffffa062b1e0>] ? SPECTRE_FBFlush+0x20/0x20 [fglrx]
[ 29.191873] [<ffffffffa050dadf>] ? IRQMGR_initialize+0x1ef/0x380 [fglrx]
[ 29.191901] [<ffffffffa050d897>] ? IRQMGR_alloc_context+0xb7/0x110 [fglrx]
[ 29.191928] [<ffffffffa050c202>] ? irqmgr_wrap_initialize+0x22/0xc0 [fglrx]
[ 29.191955] [<ffffffffa050d693>] ? firegl_irqmgr_init+0x53/0x100 [fglrx]
[ 29.191984] [<ffffffffa051e6ff>] ? hal_init_gpu+0x1cf/0x480 [fglrx]
[ 29.191988] [<ffffffff814fe9d6>] ? _raw_spin_unlock+0x16/0x40
[ 29.192012] [<ffffffffa04f8b3b>] ? firegl_open+0x2db/0x310 [fglrx]
[ 29.192031] [<ffffffffa04e71a7>] ? ip_firegl_open+0x17/0x20 [fglrx]
[ 29.192050] [<ffffffffa04e8a45>] ? firegl_stub_open+0xa5/0x110 [fglrx]
[ 29.192054] [<ffffffff8115d228>] ? chrdev_open+0xa8/0x1d0
[ 29.192057] [<ffffffff8115600f>] ? do_dentry_open+0x1df/0x2d0
[ 29.192060] [<ffffffff811640ec>] ? __inode_permission+0x5c/0xd0
[ 29.192062] [<ffffffff8115d180>] ? cdev_put+0x30/0x30
[ 29.192064] [<ffffffff811562e1>] ? finish_open+0x31/0x40
[ 29.192067] [<ffffffff81167f94>] ? do_last+0xa24/0x1160
[ 29.192069] [<ffffffff8114ba2c>] ? kfree+0x16c/0x1b0
[ 29.192089] [<ffffffffa04ea1fe>] ? KCL_MEM_SmallBufferFree+0xe/0x10 [fglrx]
[ 29.192092] [<ffffffff8116879b>] ? path_openat+0xcb/0x680
[ 29.192117] [<ffffffffa04f6ead>] ? firegl_release_helper+0x42d/0x700 [fglrx]
[ 29.192120] [<ffffffff811695a7>] ? do_filp_open+0x47/0xa0
[ 29.192122] [<ffffffff814fe9d6>] ? _raw_spin_unlock+0x16/0x40
[ 29.192125] [<ffffffff81176297>] ? __alloc_fd+0xa7/0x130
[ 29.192127] [<ffffffff81157b19>] ? do_sys_open+0x139/0x220
[ 29.192129] [<ffffffff81157c1e>] ? SyS_open+0x1e/0x20
[ 29.192132] [<ffffffff814ffd56>] ? system_call_fastpath+0x1a/0x1f

It happens 10 times in a row. Anyone else seeing this?
The graphics seem to work OK anyway, but sound was a little broken.

solaris_x84 · 10-07-2014, 02:18 PM

Quote:

Originally Posted by andy713

Because I apparently watch too much Top Gear, I thought "how hard could it be?" and "what could possibly go wrong?" and built me a new Kaveri box this week. This was current three weeks ago (with multilib), I upped the kernel version to 3.13.1 and the ATI Catalyst driver to the 14.1 beta. I've been using nvidia the last few years so I'm new to ATI. A few questions:
1. What are the sensor modules? sensors-detect didn't pick up anything.
2. Has /usr/lib64/fglrx/switchlibGL always been broken for Slackware?
3. I did get DRI working in spite of the libGL brokeness, but now google earth 7 crashes right out of the gate with signal 11. GE 6 doesn't have this problem, but it sure is slow.
4. Something in my system doesn't like b43 hardware. I got this during boot:
[ 19.100324] b43-phy0: Loading firmware version 666.2 (2011-02-23 01:15:07)
[ 19.173866] AMD-Vi: Event logged [IO_PAGE_FAULT device=00:14.4 domain=0x0014 address=0x0000000000117040 flags=0x0050]
It then proceeded to spit out many IO_PAGE_FAULT messages per second until rebooted. I had to swap out the wireless card with an Atheros I had lying around to make it stop.

Andy

Hello
I have the same problem with my AMD A8 7600 Kaveri and and Atheros based wirelesschip. I tried it with Slackware 14.1 64-bit, Mageia 4 64-Bit and Openmandriva LX 2014.1 64-Bit. For this issue exists a bugreport on redhat. With an old Knoppix 32-Bit version 7.2.0 and Kernel 3.9.6 it works. Now I have to send this APU back and use my old Richland and all work fine. I hope AMD can give us a patch.

genss · 10-07-2014, 02:41 PM

the weird temperature values are not a bug
from kernel Documentation/hwmon/k10temp

Code:

There is one temperature measurement value, available as temp1_input in
sysfs. It is measured in degrees Celsius with a resolution of 1/8th degree.
Please note that it is defined as a relative value; to quote the AMD manual:

  Tctl is the processor temperature control value, used by the platform to
  control cooling systems. Tctl is a non-physical temperature on an
  arbitrary scale measured in degrees. It does _not_ represent an actual
  physical temperature like die or case temperature. Instead, it specifies
  the processor temperature relative to the point at which the system must
  supply the maximum cooling for the processor's specified maximum case
  temperature and maximum thermal power dissipation.

The maximum value for Tctl is available in the file temp1_max.

did you say yes to some sensors-detect questions ?
on my 16h amd the sensors that returns the real cpu temperature is nct6776-isa-0290
sensors-detect does not tell me this is i just hold enter

as for broadcom, slackware comes without the firmware for b43
you need to get it otherwise and put it in /lib/firmware/

there is also some other driver that does not need that
can't remember its name, don't have a bcm card for a while now

andy713 · 10-07-2014, 09:05 PM

The broadcom wifi card worked, it just did the old "fill the syslog with IO_PAGE_FAULT messages" trick. I looked at the routine that generates this message, but I'd need to know way more about IOMMU's before I could do some non-destructive code tweaks.

I'm sure the relative temperature thing sounded like a good idea at some point, but today it says:

$ sensors
k10temp-pci-00c3
Adapter: PCI adapter
temp1: +17.9°C (high = +70.0°C)
(crit = +70.0°C, hyst = +69.0°C)

I still don't know how to interpret this.

Andy

genss · 10-07-2014, 10:00 PM

my bad

you can safely disable iommu,
even temporary as a boot parameter by adding iommu=off or iommu=soft to the lilo/grub's line at boot
the wikipedia has a good page on it

as for the temperatures
they are not for you to read as they are for bios/uefi or the kernel
they are something relative to something something (related to fan/temperature/cpu's casing and idk)

i have that sensor and another one that outputs (and one more that outputs power used)

Code:

k10temp-pci-00c3
Adapter: PCI adapter
temp1:        +19.8°C  (high = +70.0°C)
                       (crit = +70.0°C, hyst = +67.0°C)

nct6776-isa-0290
Adapter: ISA adapter
Vcore:        +0.97 V  (min =  +0.00 V, max =  +1.74 V)
in1:          +1.79 V  (min =  +0.00 V, max =  +0.00 V)  ALARM
AVCC:         +3.31 V  (min =  +0.00 V, max =  +0.00 V)  ALARM
+3.3V:        +3.31 V  (min =  +0.00 V, max =  +0.00 V)  ALARM
in4:          +0.02 V  (min =  +0.00 V, max =  +0.00 V)  ALARM
in5:          +1.67 V  (min =  +0.00 V, max =  +0.00 V)  ALARM
3VSB:         +3.44 V  (min =  +0.00 V, max =  +0.00 V)  ALARM
Vbat:         +3.38 V  (min =  +0.00 V, max =  +0.00 V)  ALARM
fan1:           0 RPM  (min =    0 RPM)  ALARM
fan2:         594 RPM  (min =    0 RPM)  ALARM
fan3:           0 RPM  (min =    0 RPM)  ALARM
SYSTIN:       +39.0°C  (high =  +0.0°C, hyst =  +0.0°C)  ALARM  sensor = thermistor
CPUTIN:       +39.5°C  (high = +80.0°C, hyst = +75.0°C)  sensor = thermistor
AUXTIN:      +127.5°C  (high = +101.0°C, hyst = +101.0°C)  ALARM  sensor = thermistor
cpu0_vid:    +1.550 V
intrusion0:  ALARM
intrusion1:  ALARM

CPUTIN is the real one

a glance at the nct6776 chips datasheet tells me it's a motherboards chip, that i guess translates the cpu's temp to real temp when read

i guess you have a different chip
note that i found that by answering yes to some sensors-detect questions, that it warned were potentially unsafe