[VANILLA-L:596] Re: Solaris SIGSEGV waitq problem solved

To: Multiple recipients of list <vanilla-l@anthrax.ecst.csuchico.edu>
Subject: [VANILLA-L:596] Re: Solaris SIGSEGV waitq problem solved
From: Michael Kantner <kantner@cds.caltech.edu>
Date: Wed, 7 May 1997 08:27:27 -0700 (PDT)
Reply-To: vanilla-l@anthrax.ecst.csuchico.edu
Sender: vanilla-l@anthrax.ecst.csuchico.edu

Hello all,

James correctly identied the cause of the Solaris waitq seg-fault.
The waitq code implicitly assumed that the shmat() call returned
identical virtual addresses.  However, that is not an OS requirement.

Two solutions were possible:
1)  Force shmat() to assign identical addresses
2)  Remove the identical address dependency

Both were simple, and 2) was implemented.  The patches have been 
tested (under Linux and Solaris) and submitted to James.  They
will be in the next release.

This bug had existed for about 2 years.  Until recently, it had
never caused a problem.

-Michael Kantner
kantner@cds.caltech.edu

p.s.  For those wondering, I introduced this bug 2 years ago when
I fixed several other bugs by rewriting the waitq code.

p.p.s.  If you want the patches before the next release, e-mail me.

Prev by Date: [VANILLA-L:597] FW: problems compiling Server2.7pl6.tar
Next by Date: [VANILLA-L:595] Re: Andrew Markiel's article
Prev by thread: [VANILLA-L:595] Re: Andrew Markiel's article
Next by thread: [VANILLA-L:597] FW: problems compiling Server2.7pl6.tar
Index(es):
- Main
- Thread