[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: New problem with an hp CD5750



On Mon, Apr 02, 2007 at 11:35:57PM +0200, jp.pozzi@izzop.net wrote:
> Somebody on the list give me an advice to switch (in the BIOS) IDE
> emulation from "Legacy IDE" to "Native" I noticed that when I select
> "Legacy IDE" the access turns to UDMA 100 with no other choice.
> When I try it I get a kernel panic at boot ... not syncing ... please
> append a correct "root=".
> So it is a misfit between the drivers and the system or with the way the
> SATA drivers works.
> Are the SATA hd devices always called HDx with the SATA driver working
> or are they called SDx like the Scsi hd ?
> On one of my home systems I lost an SCSI disk and replace it by a
> standard IDE one and the disk is seen as SDE after the SCSI disks
> (kernel 2.6.20).

All SATA drivers in linux show up as /dev/sdX.  The new libata PATA
drivers do too, meaning with new kernels (very new 2.6 series) using the
libata drivers for everything, you will have no /dev/hdX for any devices
at all.

> Knoppix is Debian based, so the system seems to be OK, and I update it
> with apt-get set to etch.
> I recompile the last 2.6.19 with SATA set to "Off" and the machine boots
> (Legacy IDE) with the same bad disk performance. A simple "vi" on a file
> is looooong as on the venerable Amiga which I used as reference.
> (Yes an Amiga 2000 works with Linux, this one has known its firts kernel
> before the 2.0 and runs with a Debian since 2.0.x, I think x was 10 to
> 12 many years ago, it runs now 2.6.18 with full success). 
> On my "big" (2x AMD 2Ghz) system the same "hdparm -t /dev/sde1" gives
> 66.46 Mo/s on the IDE disk.

Well if it is emulating an old ide controller, then you only get generic
IDE drivers, which means no DMA but rather PIO, which is limited to
about 2MB/s.  That's the speed my old 486 gets on disk accesses.

You might need a 2.6.20 or newer kernel along with setting the bios to
run the SATA in native mode in order to get anything reasonable
performance wise.

--
Len Sorensen



Reply to: