Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 17 Dec 2002 11:40:52 +0900
From:      Hidetoshi Shimokawa <simokawa@sat.t.u-tokyo.ac.jp>
To:        Sameh Ghane <sw@anthologeek.net>
Cc:        freebsd-firewire@freebsd.org, freebsd-mobile@freebsd.org
Subject:   Re: if_fwe and interrupts weirdness in current (maybe cardbus related ?)
Message-ID:  <ybsbs3l9xl7.wl@ett.sat.t.u-tokyo.ac.jp>
In-Reply-To: <20021216143510.GC14230@anthologeek.net>
References:  <20021216143510.GC14230@anthologeek.net>

next in thread | previous in thread | raw e-mail | index | archive | help
At Mon, 16 Dec 2002 15:35:10 +0100,
Sameh Ghane wrote:
> The other problem I am seeing is interrupt storm:
> 
>                     /0   /10  /20  /30  /40  /50  /60  /70  /80  /90  /100
> root     irq10: cbb XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX 
> root           idle XXXXXXXXX 
>              <idle> XXXXX 
> 
> The laptop's dmesg is attached for more info about cbb, etc.
> 
> Even when no network activity occur, top(1) shows :
> last pid:   666;  load averages:  0.02,  0.03,  0.00    up 0+01:28:15  14:25:52
> 20 processes:  1 running, 19 sleeping
> CPU states:  0.0% user,  0.0% nice,  0.0% system, 90.1% interrupt,  9.9% idle
>                                                   ^^^^^
> Mem: 7428K Active, 14M Inact, 19M Wired, 8800K Buf, 457M Free
> Swap: 1024M Total, 1024M Free
> 
> When there is no such problem on the workstation.
> 
> The only things that change from a fresh -current install are:
> 
> /boot/device.hints adds:
> # FreeBSD freezes when trying to use DMA on my SiS5591
> hw.ata.ata_dma="0"
> hw.ata.atapi_dma="0"
> # else I get a double page fault:
> hw.pcic.intr_path="1"
> hw.pcic.irq="0"
> 
> /boot/loader.conf:
> snd_ich_load="YES"
> acpi_load="YES"
> firewire_load="YES"
> if_fwe_load="YES"
> vesa_load="YES"
> boot_verbose="YES"
> linux_load="YES"
> nvidia_load="YES"
> 
> I tried to add the DEVICE_POLLING option in order to reduce the interrupt usage,
> it did not change *anything* wrt this bug.
> 
> It does not seem to come from the fwe interface. Rather from the bus or
> something.
> 
> I would be happy to help find out a solution to these problems.

Are you sure the device polling was really enabled?

To enable it, you need to do 'sysctl kern.polling.enable=1',
'ifconfig fwe0 down;ifconfig fwe0 up' and make sure you get
'fwohci0: fwochi_set_intr: 0' in your dmesg.

I have similar problem on my laptop. It turns out that ACPI event
handler shares the IRQ(it cannot read from dmesg) and it consumes much
of the interrupt time.
Could you try to disable ACPI?

/\ Hidetoshi Shimokawa
\/  simokawa@sat.t.u-tokyo.ac.jp
PGP public key: http://www.sat.t.u-tokyo.ac.jp/~simokawa/pgp.html

To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-firewire" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?ybsbs3l9xl7.wl>