[prev in list] [next in list] [prev in thread] [next in thread] 

List:       linux1394-devel
Subject:    Re: [PATCH 2.6.18-rc6-mm1] ieee1394: sbp2: slightly reorder
From:       Stefan Richter <stefanr () s5r6 ! in-berlin ! de>
Date:       2006-09-12 11:04:38
Message-ID: 45069446.5080401 () s5r6 ! in-berlin ! de
[Download RAW message or body]

will wrote:
...
> You were quite right about the use of SUCCESS / FAILURE with the handlers, I
> miss read the code sorry about that.

No problem, I'm actually glad you made me check this again. (BTW, sites
like http://lxr.free-electrons.com/ are nice to browse the code. Cscope/
KScope/ Cbrowser can do that too locally but I didn't try them myself yet.)

...
> After the failure has happened the module has a lot of trouble
> unloading, it tries to issue an Xwrite which then eventually aborts, and
> retries this a number of times. This doesn't seem to help much as the driver
> is unable to reconnect when it reloads. I guess I probably need to do the
> same thing with the SCSI drivers / layer etc.

I think sr_mod issues a lot of such commands even after the device was
already offlined by scsi_mod. Quite a nuisance. I haven't looked further
into it yet and probably won't do so anytime soon.

...
> do you think that there is a more aggressive
> error recovery strategy which could be applied if the 1st one fails (making
> use of the escalating SCSI error handling routines). For example sort of
> re-initialisation code?

We should try the carrot before the stick. If the Test Unit Ready after
fetch agent reset still fails, one thing that could be tried is to
insert a pause of, say, half a second after the agent reset.

Going the more aggressive route, one thing that could be considered for
sbp2scsi_reset() is the Target Reset task management function as
described in SBP-2 clause 10.4.4. This needs additional consideration
for targets with multiple logical units though. And as far as semantics
as per the spec go, Target Reset won't really help more than Fetch Agent
Reset for this case. But who knows the inner workings of real target
implementations.

I have no advice on escalation of error handling in SCSI core and the
CD-ROM driver; I don't know them well enough.

> When I respond to mails should I reply directly to you or straight to
> linux1394-devel@lists.sourceforge.net or cc the list etc?

Please respond to the list, unless you have big logs to send or other
information that is not suitable for the list. (You need to mark the
latter explicitly, else I might accidently quote it on the list.) It is
customary on linux-kernel related lists to keep every responder of a
thread in Cc. But I will of course get mails to the list without Cc too.
-- 
Stefan Richter
-=====-=-==- =--= -==--
http://arcgraph.de/sr/

-------------------------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
mailing list linux1394-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux1394-devel
[prev in list] [next in list] [prev in thread] [next in thread]