[6131] in linux-scsi channel archive
Re: Got it! SG select() bug evidence for 2.2.2
daemon@ATHENA.MIT.EDU (Gilbert, Douglas)
Fri Mar 19 11:47:57 1999
From: "Gilbert, Douglas" <douglas.gilbert@rbcds.com>
To: "'linux-scsi@vger.rutgers.edu'" <linux-scsi@vger.rutgers.edu>,
"'cdwrite@lists.debian.org'" <cdwrite@lists.debian.org>
Cc: "'xiphmont@mit.edu'" <xiphmont@mit.edu>
Date: Fri, 19 Mar 1999 11:32:23 -0500
This message is in MIME format. Since your mail reader does not understand
this format, some or all of this message may not be legible.
------_=_NextPart_001_01BE7226.11BABACC
Content-Type: text/plain;
charset="ISO-8859-1"
Monty,
[snipped the original, will answer snippets]
> "SCSI transport: writeable when reading packet"
> ...
> ... because I'm only 80% certain it's SG.
Well, make that 99.9% in sg (and cdparanoia :-)) in
2.2.2-ac6+7 and 2.2.3-ac1 . It is "fixed" in
2.2.3-ac2+3. Joerg's patches have probably picked that
up as well.
The select() code addition you describe in cdparanoia
alpha 9.4 has the side effect of distinguishing between
a sg driver that supports command queuing (desirable)
and one that doesn't (the original).
Your post also explains to me why most versions of
cdparanoia worked, but some failed.
The "fix" was to turn turn off command queing by default :-(
The following analysis with strace in your post is just
what I would expect. As to your comment about "closing an
SG fd with a command oustanding is a great way to cause to
'oops'", my driver improves this situation but doesn't
remove it. My driver is still susceptible to a rapidly
following "rmmod sg" as I have reported before to this
newsgroup. The SIG_BLOCK is not necessary with my driver.
Is there any possiblity of changing that select() code
in cdparanoia alpha 9.4 ??
In summary: there is no serious bug in the poll() area of
2.2.x kernels exposed here.
Doug Gilbert
------_=_NextPart_001_01BE7226.11BABACC
Content-Type: text/html;
charset="ISO-8859-1"
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2//EN">
<HTML>
<HEAD>
<META HTTP-EQUIV="Content-Type" CONTENT="text/html; charset=ISO-8859-1">
<META NAME="Generator" CONTENT="MS Exchange Server version 5.5.2232.0">
<TITLE>Re: Got it! SG select() bug evidence for 2.2.2</TITLE>
</HEAD>
<BODY>
<P><FONT SIZE=2>Monty,</FONT>
<BR><FONT SIZE=2>[snipped the original, will answer snippets]</FONT>
</P>
<P><FONT SIZE=2>> "SCSI transport: writeable when reading packet"</FONT>
<BR><FONT SIZE=2>> ...</FONT>
<BR><FONT SIZE=2>> ... because I'm only 80% certain it's SG.</FONT>
</P>
<P><FONT SIZE=2>Well, make that 99.9% in sg (and cdparanoia :-)) in </FONT>
<BR><FONT SIZE=2>2.2.2-ac6+7 and 2.2.3-ac1 . It is "fixed" in </FONT>
<BR><FONT SIZE=2>2.2.3-ac2+3. Joerg's patches have probably picked that</FONT>
<BR><FONT SIZE=2>up as well.</FONT>
</P>
<P><FONT SIZE=2>The select() code addition you describe in cdparanoia</FONT>
<BR><FONT SIZE=2>alpha 9.4 has the side effect of distinguishing between</FONT>
<BR><FONT SIZE=2>a sg driver that supports command queuing (desirable)</FONT>
<BR><FONT SIZE=2>and one that doesn't (the original).</FONT>
</P>
<P><FONT SIZE=2>Your post also explains to me why most versions of</FONT>
<BR><FONT SIZE=2>cdparanoia worked, but some failed.</FONT>
</P>
<P><FONT SIZE=2>The "fix" was to turn turn off command queing by default :-(</FONT>
</P>
<P><FONT SIZE=2>The following analysis with strace in your post is just</FONT>
<BR><FONT SIZE=2>what I would expect. As to your comment about "closing an</FONT>
<BR><FONT SIZE=2>SG fd with a command oustanding is a great way to cause to</FONT>
<BR><FONT SIZE=2>'oops'", my driver improves this situation but doesn't</FONT>
<BR><FONT SIZE=2>remove it. My driver is still susceptible to a rapidly</FONT>
<BR><FONT SIZE=2>following "rmmod sg" as I have reported before to this</FONT>
<BR><FONT SIZE=2>newsgroup. The SIG_BLOCK is not necessary with my driver.</FONT>
</P>
<P><FONT SIZE=2>Is there any possiblity of changing that select() code</FONT>
<BR><FONT SIZE=2>in cdparanoia alpha 9.4 ??</FONT>
</P>
<BR>
<P><FONT SIZE=2>In summary: there is no serious bug in the poll() area of</FONT>
<BR><FONT SIZE=2>2.2.x kernels exposed here.</FONT>
</P>
<BR>
<P><FONT SIZE=2>Doug Gilbert</FONT>
</P>
</BODY>
</HTML>
------_=_NextPart_001_01BE7226.11BABACC--
-
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@vger.rutgers.edu
