Re: DMA issue in 2.6.12 ???
On Wed, Aug 17, 2005 at 01:06:30PM -0500, Ardo van Rangelrooij wrote:
> I ran into an interesting DMA issue in 2.6.12 vs. 2.6.11:
> First some info on the system I'm running the latest and greatest sid on:
> Asus A8V Deluxe
> AMD Athlon64 3000+
> 1G dual channel DDR (2x512MB)
> /dev/sda = Maxtor 6B200M0
> /dev/hda = Sony DVD-ROM DDU1622
> /dev/hdb = WDC WD1200JB-75CRA0
> /dev/hdc = Maxtor 6B200R0
> /dev/hdd = WDC WD1200JB-00EVA0
> /dev/sda has / and swap
> /dev/hdb and /dev/hdd are in /dev/md0 as raid-1
> When using 2.6.11-9 everything is ok and as expected: The DMA flags are set
> when running `cat /proc/ide/hd[bcd]/settings`, running `hdparm -T` gives
> 1800+ MB/sec, and running `hdparm -t` gives 45MB/sec for /dev/md0 and 54MB/sec
> for /dev/hdc. So far so good.
> However, when running 2.6.12-1 things are not as expected: The DMA flags are
> not set, and running hdparm I get much lower numbers. And I get
> > hdparm -d 1 /dev/hdc
> setting using_dma to 1 (on)
> HDIO_SET_DMA failed: Operation not permitted
> using_dma = 0 (off)
> When booting nothing in the log files indicates something's wrong or whatever.
> Anybody any thoughts?
Somehow the new kernel is loading ide-generic first and it takes control
of the ide interfaces and of course doesn't allow dma at all since that
is a chipset specific feature.
lsmod should show which driver has how many uses of it. dmesg should
also show which driver grabbed each device.
the initrd tools has loaded ide-generic too early in the past, as has
some driver dependancy bugs once upon a time too.