Date: Tue, 9 Apr 2002 23:50:20 +0200 From: Thomas Quinot <thomas@cuivre.fr.eu.org> To: freebsd-scsi@freebsd.org Subject: Re: xpt_bus_deregister Message-ID: <20020409235020.A2397@melusine.cuivre.fr.eu.org> In-Reply-To: <20020409094509.A4569@melusine.cuivre.fr.eu.org>; from thomas@melusine.cuivre.fr.eu.org on Tue, Apr 09, 2002 at 09:45:09AM %2B0200 References: <20020409094509.A4569@melusine.cuivre.fr.eu.org>
next in thread | previous in thread | raw e-mail | index | archive | help
Le 2002-04-09, Thomas Quinot écrivait :
> problem with xpt_bus_deregister. Apparently, there is one reference
> to the bus which is counted but never released. As a consequence,
> a call to xpt_bus_deregister(cam_sim_path(scp->sim)) will not actually
> remove the bus from xpt_busses, and of course a subsequent cam_sim_free
> destroys the SIM with extreme prejudice.
>
> Any idea where I should start looking for this reference leak?
OK, I think I found it. When attaching the SIM, I setup an
async callback on AC_LOST_DEVICE. This callback counts as
a reference to the wildcard device, which prevents the release
of the wildcard target, and therefore of the bus. Disabling
the callback before calling xpt_bus_deregister & al. resolved
the problem.
I suspect that this problem is dormant in many other SCSI
host adapter drivers. Do others have experience with removing
a SIM from a running system?
Thomas.
--
Thomas.Quinot@Cuivre.FR.EU.ORG
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-scsi" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20020409235020.A2397>