Skip site navigation (1)Skip section navigation (2)
Date:      Mon, 20 Jun 2005 11:43:27 -0300 (EST)
From:      <scuba@centroin.com.br>
To:        Bruce Evans <bde@zeta.org.au>
Cc:        freebsd-performance@FreeBSD.org
Subject:   Re: Slave IDE HDD not working in UDMA5
Message-ID:  <Pine.BSI.4.33.0506201138550.17127-100000@hypselo.centroin.com.br>
In-Reply-To: <20050620222051.T13863@delplex.bde.org>

next in thread | previous in thread | raw e-mail | index | archive | help
On Mon, 20 Jun 2005, Bruce Evans wrote:

|On Thu, 16 Jun 2005 scuba@centroin.com.br wrote:
|
|> =09This was first posted on freebsd-question, but I could not find
|> the solution yet.
|> =09Maybe you could help me.
|
|Maybe a little.  I don't use 5.4...

=09It happens on 5.3 also. Not tested on 4.x.

|> =09I=B4ve installed FreeBSD 5.4 on a new machine with the following
|> hardware:
|>
|> =09Asus P4P800 SE  (BIOS v. 1008)
|> =092GB RAM ( 4x 512 DDR400 )
|> =092 HDD Samsung SP0802N (80GB 7200rpm ata-100) 80 pins cable.
|>
|> =09The HD were formated with newfs defaults, and the following
|> results were the same using both as master (primary e secondary) or with=
 a
|> master / slave (same interface).
|>
|> =09With diskinfo both performance are the same, but with "dd", the
|> second disc (the slave or the secondary master), is always worst as if i=
t
|> were working in DMA2.
|
|I remember a commit to the ata driver to fix misprogramming of DMA timing
|on an Intel chipset for devices and/or channels other than the first.  I'm
|not sure if 5.4 has the bug or the fix.

=09It seems that the bug is still there.

|diskinfo only tests reading, and you only showed a dd test using writing
|(to a file), so the problem is apparently only that writing to the second
|drive is slow.

=09Right.

|> =09what should be the right results?
|
|Swap the devices to see if it is a drive problem (unlikely with the
|same model of drive but...  I have one system that apparently has
|worse timing on the secondary channel.  This showed up as writes
|causing subsequent reads to be slow -- apparently the writes caused
|some errors and error correction as perfect except for slowing things
|down).

=09I=B4d already did it. It=B4s not a hardware problem. Always happen on
the second drive, no matter which is it.


- Marcelo





Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSI.4.33.0506201138550.17127-100000>