[linux-audio-user] 2.6.18-rt5

Yves Potin yves at jazzcomputer.org
Sat Sep 30 08:42:19 EDT 2006


        Hi. 
        I have achieved booting a 2.6.18-rt5 on a dual Opteron X2
system. As I've just had the idea of disabling HPET with this version of
the patch, I can't say if it could work on an earlier version (it hanged
on all versions theretofore). HPET has always been activated in the kernel
I've compiled before. 
        The config is there, it's only made for an AMD 8111 motherboard
with Sil SATA chipset, but IDE is activated without any need of initrd :
http://ypotin.nerim.net/myconfig 
        Apparately, the rtc works better. 
        This kind of messages (2.6.17.7) : 

Sep 25 10:43:42 fairlight kernel: softirq-tasklet/20[CPU#1]: BUG in
__tasklet_action at kernel/softirq.c:493
Sep 25 10:43:42 fairlight kernel:  [<c011e467>] __WARN_ON+0x67/0x90 (8)
Sep 25 10:43:42 fairlight kernel:  [<c0122d0b>] __tasklet_action+0xeb/0xf0
(48)
Sep 25 10:43:42 fairlight kernel:  [<c01234f6>] ksoftirqd+0xf6/0x1a0 (24)
Sep 25 10:43:42 fairlight kernel:  [<c0123400>] ksoftirqd+0x0/0x1a0 (48)
Sep 25 10:43:42 fairlight kernel:  [<c01318bb>] kthread+0xab/0xe0 (4)
Sep 25 10:43:42 fairlight kernel:  [<c0131810>] kthread+0x0/0xe0 (12)
Sep 25 10:43:42 fairlight kernel:  [<c0100e65>]
kernel_thread_helper+0x5/0x10 (1
6)

        And also that (I had tons of that) : 

Sep 28 21:35:13 fairlight kernel: bug in rtc_read(): called in state
S_IDLE!
Sep 28 21:35:13 fairlight kernel: Read missed before next interrupt
Sep 28 21:35:48 fairlight kernel: bug in rtc_read(): called in state
S_IDLE!

        All this *seems* to have disappear, after three hours using the
kernel. The rtc have stopped saying « eek » sometimes in the logs.
        Instead, I have tons of this kind of things : 

Sep 30 13:08:09 fairlight kernel: rtc: lost some interrupts at 1024Hz.
Sep 30 13:08:31 fairlight last message repeated 1899 times
Sep 30 13:09:01 fairlight kernel: 1024Hz.
Sep 30 13:09:01 fairlight kernel: rtc: lost some interrupts at 1024Hz.
Sep 30 13:09:16 fairlight last message repeated 1545 times
[...]
Sep 30 14:20:00 fairlight kernel: rtc: lost some interrupts at 1024Hz.
Sep 30 14:20:31 fairlight last message repeated 2346 times
Sep 30 14:21:32 fairlight last message repeated 3067 times
Sep 30 14:21:54 fairlight last message repeated 1148 times

         IRQ balancing is activated, and the litlle redhat's software for
that is installed and running. I have chosen to compile the enhanced rtc
in module, and have disabled its graphs because it flooded my logs without
any use for me.
        I have two questions : 
        - Could someone tell me, more clearly than in the kernel help :),
what's the use of this HPET beast  ?
        - Does it matter to disable it for real time audio purposes ?
        Thanks in advance for any enlightenment. 
        Cheers, 

Y.





P.S. : /proc/interrupts : 

           CPU0       CPU1       CPU2       CPU3       
  0:    3237178    3211763    3204014    3202106  IO-APIC
[........N/  0]-edge     timer
  6:          2          0          0          1  IO-APIC
[........./  1]-edge     floppy
  8:          0          0          0          0  IO-APIC
[........./  0]-edge     rtc
  9:          0          0          0          0  IO-APIC
[........./  0]-level    acpi
 14:      89780     107867     194304       1819  IO-APIC
[........./  0]-edge     ide0
 15:         31          0          1          1  IO-APIC
[........./  0]-edge     ide1
 16:      24128      57874      47830      32380  IO-APIC
[........./  0]-level    libata
 17:      27659          0          0          1  IO-APIC
[........./  0]-level    eth0
 18:      17715      30803      28025      12471  IO-APIC
[........./  0]-level    ohci_hcd:usb1, ohci_hcd:usb2, ohci1394
 19:     991430    1208684     744999    1458384  IO-APIC
[........./  0]-level    hdsp
NMI:          0          0          0          0 
LOC:   18863126   17966846   18728772   17747659 
ERR:          0
MIS:          0






More information about the Linux-audio-user mailing list