Re: hdparm: invalidate: busy buffer
On Mon, Dec 09, 2002 at 01:49:06PM -0800, nate wrote:
> Pigeon said:
> > Right, I've heard lots of people moaning about the RAID version. OK,
> > thanks.
> > When you say 'buggy', do you mean losing data at all? I've never had
> > anything like that, it's just not very quick.
> It's been 3 years, so my experiences are not up to date :) but when
> I had problems(it was on a new Asus CUV4X at the time) it involved the
> disk operating in PIO mode and load average jumping to 10 for many small
> tasks running, making the system unusuable. I was able to "force" the disk
> to DMA mode but a little while later(random time elapse, usually a few hours),
> there would be a bunch of DMA errors and the kernel would revert to PIO
> mode again.
> I wouldn't care so much, disk speed isn't that important to me, but when
> the load jumps to 8-10+ it becomes a problem! I mean it could take 30
> seconds to run 'top' before I got any results back
> as far as I remember I never had any data loss, nor crashes or anything
> just obscene load averages, and the inability to reliably put the drive
> in DMA mode. Using the promise controller immediately fixed it. And i've
> been using that controller in several systems ever since. I read a bunch
> of other reports at the time about problems with VIA chipsets(my older
> VIA MVP3 based K6-3 works wonderfully in DMA mode) so I knew it wasn't
> just me.
Ah, no, I've never had anything like that happen. It stays in DMA
mode, it's just slow. I haven't noticed any errors apart from the
abovementioned "invalidate: buffer busy" from hdparm. On the other hand, I
guess from your posts that your systems are far more heavily used than
3 years is roughly the age of my ECS MB. I've heard of VIA chipset
problems - AFTER I got the MB! - but got the impression that my
82C596B was one of the trouble-free variants. Maybe "trouble-free VIA
chipset" is one of those phrases like "working Windoze program".