Re: Kernel panic when re-inserting Adaptec PCMCIA card

[Alex Davis]

--- Alex Davis <alex14641@xxxxxxxxx> wrote:

> --- Alex Davis <alex14641@xxxxxxxxx> wrote:
> 
> > --- Chuck Ebbert <76306.1226@xxxxxxxxxxxxxx> wrote:
> > 
> > > In-Reply-To: <20060614022139.21737.qmail@xxxxxxxxxxxxxxxxxxxxxxx>
> > > 
> > > On Tue, 13 Jun 2006 19:21:39 -0700, Alex Davis wrote:
> > > 
> > > Same panic occurs in 2.6.17rc6:
> > > 
> > > > Jun 13 17:50:36 siafu kernel: [4295220.230000] pccard: PCMCIA card inserted into slot 0
> > > > Jun 13 17:50:36 siafu kernel: [4295220.230000] pcmcia: registering new device pcmcia0.0
> > > > Jun 13 17:50:37 siafu kernel: [4295220.281000] aha152x: resetting bus...
> > > > Jun 13 17:50:37 siafu kernel: [4295220.637000] aha152x13: vital data: rev=1, io=0xd340
> > > > (0xd340/0xd340), irq=3, scsiid=7, reconnect=enabled,
> > > >  parity=enabled, synchronous=enabled, delay=100, extended translation=disabled
> > > > Jun 13 17:50:37 siafu kernel: [4295220.637000] aha152x13: trying software interrupt, ok.
> > > > Jun 13 17:50:37 siafu kernel: [4295221.638000] scsi13 : Adaptec 152x SCSI driver;
> $Revision:
> > > 2.7 $
> > > > Jun 13 17:50:37 siafu kernel: [4295221.650000]
> > > > Jun 13 17:50:37 siafu kernel: [4295221.650000] aha152x22856: bottom-half already running!?
> > > > Jun 13 17:50:37 siafu kernel: [4295221.650000]
> > > > Jun 13 17:50:37 siafu kernel: [4295221.650000] queue status:
> > > > Jun 13 17:50:37 siafu kernel: [4295221.650000] issue_SC:
> > > > Jun 13 17:50:37 siafu kernel: [4295221.650000] current_SC:
> > > > Jun 13 17:50:37 siafu kernel: [4295221.650000] BUG: unable to handle kernel paging request
> > at
> > > > virtual address 00020016
> > > 
> > > Something is going very wrong here.  At time .637 it says it is
> > > adapter number 13 (aha152x13.)  Then at .650 it thinks it's
> > > adapter nr. 22856 (!)  Looks like some kind of pointer to the
> > > hostdata is corrupted.
> > > 
> > > Can you rmmod the driver after removing the card and see if that
> > > helps?
> > > 
> > It turns out I was trying to remove the driver before doing 'pccardctl eject'.
> > 
> > It seems that removing the driver after ejecting make the problem go away:
> > I ejected and re-inserted the card six times with no crash.
> > 
> > I'll continue testing just to make sure.

I now know what the problem is.

In drivers/scsi/aha152x.c, there is an array of two scsi hosts:
        static struct Scsi_Host *aha152x_host[2];
There is also the aha152x_probe_one method.

The problem is that each re-insert of the card creates a new 'host' via aha152x_probe_one,
which increments the registered_count variable. This variable is used as an index into the afore-
mentioned 2-element host array. When registered_count is 2 or greater we're in no-man's land,
hence the crashes. I'm currently testing a fix for this, which I hope to post soon.

-Alex

I code, therefore I am

__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/