lkml.org 
[lkml]   [2002]   [Sep]   [6]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: One more bio for for floppy users in 2.5.33..

On Fri, 6 Sep 2002, Helge Hafting wrote:
>
> I can think of one case where large readahead hurts for floppy, even
> with partial completion:
>
> 1. Grab a stack of floppies
> 2. Try mounting (or mount+ls) one after another,
> in search of the right one.

Note that the delay for motor on/off is _much_ larger than the actual
delay for seeking.

The seek itself is on the order of a few ms, with the head settle time
being in the tens (possibly even a few hundred) ms per track. So assuming
you end up reading 4 tracks or so due to readahead, that's still in the
range of about one second.

In contrast, the motor on/off time is something like 5 seconds if I
remember correctly. Of course, you can certainly eject the floppy while
the motor is still running, but I'd suggest against it.

> So I think a smaller readahead might make sense for floppies,
> unless people don't do this sort of search anymore. I don't.

I do agree that a 64kB readahead is likely to be excessive on a floppy,
I'm just saying that I doubt it will be all that noticeable in most cases.

The absolute worst case is when opening, reading a sector, and closing
again several times in succession, at which point right now we'll end up
serializing. But even at 64kB, that's going to be faster than most people
can change floppies if they actually want to even glance at what the
contents are.

The reason I want the first sectors to be returned early is that I thought
it was quite noticeable to do just a simple "ls" on the floppy when I
tested. Of course, that may be just me: it's literally been several years
since I really used floppies, and maybe they really always were that slow.
But I thought the root directory was on track zero, so it _should_ return
it first thing.

Oh, well. I don't seem to be the only one who doesn't use the dang things
any more. The floppy driver has been broken in 2.5.x for half a year or
whatever, and there weren't _that_ many people who ever even mentioned it.

Linus

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2005-03-22 13:28    [W:0.083 / U:0.232 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site