Linux ate my DOS! (swap problem I think)

Linux ate my DOS! (swap problem I think)

Post by Jason LaPier » Sun, 27 Jun 1993 03:28:58



  This is probably my fault but I have no idea what I could have done
to cause it. My fstab has the swap partition set to  /dev/hda3. I have
been happily linuxing away for a couple weeks (on my most recent
installation), when on a recent reboot I received a message from swapon
that /dev/hda3 had an invalid swap signature. So I did a mkswap on the
partition again. All is fine and good. Several days later I tried to mount
my dos partition (/dev/hda1) and was promptly informed that it was the
wrong file type. On trying to boot this partition, the destruction of DOS
was confirmed (not necessarily a bad thing I know).

  I then remembered noticing that in the /dev directory there was a link
to /dev/hda1 called swap. I had never messed with it before so I chose to
leave it. (This installation was from SLS 1.02) Was this my critical
error? Should /dev/swap be a link to whatever the swap partition is?
If so, why doesn't this cause problems right away, and why didn't it ever
do this before? Inquiring minds would like to know 8^)

  Also, I've pretty much accepted the fact that I'm going to have to rebuild
the DOS partition. But before I start I'd really appreciate it if someone
could confirm my guess at the problem. I don't really use DOS often enough
to be positive that the two events (swap failure and DOS failure) were
actually related.

  Any and all advice appreciated.

        -Jason-

Jason LaPierre
Univerisity of Rochester

 
 
 

Linux ate my DOS! (swap problem I think)

Post by Eric J. Schwertfeg » Wed, 30 Jun 1993 01:14:59


        There have been several seemingly contradictory messages
lately concerning HD throughput.  I'd like to find out a few
facts before I make a decision on what disk system to buy for my
computer.

        There seems to be a great deal of debate as to the speed
of SCSI hard disks under Linux.  The debate seems to center over
two ideas.  First is that the bus-mastering disk controllers
suffer a severe penalty due to the need to map logical addresses
to physical addresses.  Since the 386 and up usually use memory
blocks of 4K in size, I don't think that the overhead for this
address translation would seriously impact the throughput.  Even
with a disk and controller capable of 5MB/sec, you'd have to
take 100 microseconds, or at least 2500 clock cycles, to slow
the throughput down by more than 10%.

        On the other hand, a few people seem to be under the
impression that while an IDE disk drive is reading sectors into
its buffer, the CPU is busy-waiting.  Again, this doesn't make
much sense, since the disk can trigger an interrupt as soon as
the entire sector (or group of sectors, if the driver supports
multi-sector R/W) is read, and then the CPU can transfer
buffer->RAM, at the full speed of the bus (full speed for a R/W
operation, admittedly less than busmastering throughput).

        There are several other theoretical enhancements to IDE
throughput, but there's not much use talking theoretical at the
moment.  Unless someone convinces me that it's worth an extra
$200 dollars for me to get an AHA1542 or Ultrastore 34F, I plan
on going with IDE, just to investigate those theoretical
enhancements.

        I guess in order to find out the truth of the matter, I
need to get some "benchmark" information from people that are
actually using Linux on similar machines.  What I'd like to get
is a few facts and figures, mostly IOZone's and kernal compile
times, including such information as compile options (ie, SCSI
support, TCP/IP support, etc), and system configurations
(CPU, memory, hard disk & controller types, etc).  I'd prefer
mailing, and I'll summarize, if there's interest.

        I'd also be interested in talking to anyone that has
played with the IDE multi-sector R/W ability, or more unlikely,
the DMA ability (a $14 busmastering controller?  I LIKE that
idea :-)

--


 
 
 

Linux ate my DOS! (swap problem I think)

Post by Mark A. Dav » Wed, 30 Jun 1993 03:30:37



Quote:>    There have been several seemingly contradictory messages
>lately concerning HD throughput.  I'd like to find out a few
>facts before I make a decision on what disk system to buy for my
>computer.

[...]

Quote:>    There are several other theoretical enhancements to IDE
>throughput, but there's not much use talking theoretical at the
>moment.  Unless someone convinces me that it's worth an extra
>$200 dollars for me to get an AHA1542 or Ultrastore 34F, I plan
>on going with IDE, just to investigate those theoretical
>enhancements.

There is a lot more to the issue than just speed.  I tend to think they are
about the same speed (but not if you compare SCSI/2 fast...).

You should consider that SCSI is a real and solid standard.  You can get
SCSI QIC drives, 8mm drives, DAT drives, Hard Drives, CD-ROM drives, optical
drives, etc......  IDE is extremely limiting.  In addition, since IDE is
STILL a throwback to clone stuff, you can only use 2 drives on it.  With SCSI,
you can have up to 7 devices on each channel (and with more with additional
channels).

SCSI is really, still the only standard which comes close to a "universal"
standard for any kind of drive technology.  It can also be used with things
as strange a printers, networking, and serial cluster units.  You may not
need the capability now- but $ well spent now can save headaches later...

--
  /--------------------------------------------------------------------------\
  | Mark A. Davis    | Lake Taylor Hospital | Norfolk, VA (804)-461-5001x431 |

  \--------------------------------------------------------------------------/

 
 
 

Linux ate my DOS! (swap problem I think)

Post by Philip Danie » Wed, 30 Jun 1993 19:05:26



[SCSI vs IDE]

Quote:>SCSI is really, still the only standard which comes close to a "universal"
>standard for any kind of drive technology.  It can also be used with things
>as strange a printers, networking, and serial cluster units.  You may not
>need the capability now- but $ well spent now can save headaches later...

  Consider that if you buy a big IDE harddrive you are stuck with it, whereas
if you buy a SCSI disk you can move it on to another computer (say a PowerPC
or MIPS box if they ever become cheap enough :( ).

--


Brain failure (cerebral coretex dumped)

 
 
 

Linux ate my DOS! (swap problem I think)

Post by Eric Jeschk » Thu, 01 Jul 1993 07:19:43


Check out the article in the June 19, 93 PC Magazine (V12 N12),
page 219 "Fast and Flexible: SCSI Technology Grows Up".  It has
some good arguments for SCSI vs. IDE.

--
Eric Jeschke                      |          Indiana University


 
 
 

1. Badblock on swap, I think; mkswap really done?

I am rebuilding a kernel. My 386 /16 laptop usually takes 3 1/2 hours for
this, however, I commited the whole drive (61 megs, whew!) to linux this
time. I know there is a bad block in there where DOS used to reside
without problems. This is where the swap partition is now (5.5 megs).
 question: this rebuild is in its ninth hour. Do you think the swap
partition, though it claims to be on, is not? How can I tell the swap
partition to ignore that block in future? After the rebuild, can I delete
that swap partition and rewrite it without dealing with another rebuild?

Thank-you
Dennis C

2. CDE dtfile, Solaris 2.5.1

3. Odd problem: Linux tries to eat DOS

4. SMP machines, what to watch for?

5. Problem with swap space being eaten up on 2.4 ?

6. X problem?

7. Will DOS Format Command Eat Linux Partition?

8. Xfree86/GX1

9. BJ200 Eats 3 Last Lines In Linux, Not Dos

10. linux ate my dos partition....

11. Linux eats Dos on harddisk > 1024 cyl

12. HELP, Dos ate my LINUX partition.

13. DOS, OS2, os2 bootmanager, linux native and linux swap in one d