[prev in list] [next in list] [prev in thread] [next in thread]
List: hpux-devtools
Subject: RE: HPUX-DEVTOOLS: Important: Enquiry about "panicstr" (fwd)
From: "Abhijit Toley" <abhijit_toley () symantec ! com>
Date: 2006-10-31 11:33:03
Message-ID: 0ED7969A003F1B48BE956A4CA691B8010281E0A5 () ITPXCHCLN1 ! enterprise ! veritas ! com
[Download RAW message or body]
Hi Sabya,
Well, this is not new code. And we were wondering if
its too bad to have a dependency on "panicstr". But as
you said, already there are a lot of dependencies on
"panicstr". So lets have one more :) unless its really
really bad (which it doesn't seem to be).
Yes, if we develop new code, we would use what you have
suggested.
Does that sound good?
Thanks a lot,
Abhijit.
> -----Original Message-----
> From: owner-hpux-devtools@cxx.cup.hp.com
> [mailto:owner-hpux-devtools@cxx.cup.hp.com] On Behalf Of
> Sabyasachi Sengupta
> Sent: Tuesday, October 31, 2006 1:59 PM
> To: Abhijit Toley
> Cc: hpux-devtools@cxx.cup.hp.com; Prashant Gupta; Amol Katkar
> Subject: RE: HPUX-DEVTOOLS: Important: Enquiry about "panicstr" (fwd)
>
>
>
> Hi Abhijit,
>
> Filesystem is synced only in cases of panics, but not in
> cases of MCA/INIT. You could subscribe to
> gsp_shutdown_state_t = GSP_EARLY_SHUTDOWN for panics, which
> is called pretty much early in the panic processing (in fact
> much before "panicstr" is set - there are many things done in
> between calling the early callbacks and setting panicstr). I
> don't think its a good idea to have a close dependency on
> panicstr (though there are many dependencies on it, already).
> Just a suggestion if you are developing new code. For panics,
> your gsp_shutdown_type_t should be GSP_PANIC during registration.
>
> For MCA/INIT events, you could subscribe to
> gsp_shutdown_state_t = GSP_CRASH. This doesn't sync
> filesystems. For MCA/INIT registration, you need to specify
> GSP_MCA|GSP_INIT as mask in your call.
>
> Thanks,
> Sabya
>
> --
>
> HPUX Kernel debugger tools project
> Ph: 91-80-22051892
> Mailstop: IE 2 Floor, Blue Wing-2, Cube 31
>
> On Tue, 31 Oct 2006, Abhijit Toley wrote:
>
> + Date: Tue, 31 Oct 2006 13:43:19 +0530
> + From: Abhijit Toley <abhijit_toley@symantec.com>
> + To: Sabyasachi Sengupta <sabyasg@hp.com>
> + Cc: hpux-devtools@cxx.cup.hp.com,
> + Prashant Gupta <Prashant_Gupta@symantec.com>,
> + Amol Katkar <amol_katkar@symantec.com>
> + Subject: RE: HPUX-DEVTOOLS: Important: Enquiry about
> "panicstr" (fwd)
> +
> +
> + Thanks Sabya.
> +
> + "panicstr" is set before the filesystems are synced.
> + The file system syncing can be a huge time (order of
> minutes). We do
> + not want to wait for minutes. But we can wait for a few seconds for
> + the "panicstr" to get set.
> +
> + Regards,
> + Abhijit.
> +
> +
> +
> + > -----Original Message-----
> + > From: owner-hpux-devtools@cxx.cup.hp.com
> + > [mailto:owner-hpux-devtools@cxx.cup.hp.com] On Behalf Of
> Sabyasachi
> + > Sengupta
> + > Sent: Tuesday, October 31, 2006 1:07 PM
> + > To: Abhijit Toley
> + > Cc: hpux-devtools@cxx.cup.hp.com; Prashant Gupta; Amol Katkar
> + > Subject: Re: HPUX-DEVTOOLS: Important: Enquiry about "panicstr"
> + > (fwd)
> + >
> + >
> + > Hi,
> + >
> + > I saw that my response to hpux-devtools and to Semantec
> folks were
> + > bouncing back. Attempting a resend. HPUX 11.23 has implementation
> + > for subsystem shutdown notification as I had indicated in
> the emails
> + > below.
> + >
> + > Stan: I am reading your suggestions on
> + > panic_callback_ok_to_resume_panic (which is not yet available in
> + > HPUX today), HPUX doesn't handle failures from crash / shutdown
> + > callbacks as yet, I will pass on the input to the
> appropriate team.
> + >
> + > Thanks,
> + > Sabya
> + >
> + > --
> + >
> + > HPUX Kernel debugger tools project
> + > Ph: 91-80-22051892
> + > Mailstop: IE 2 Floor, Blue Wing-2, Cube 31
> + >
> + > ---------- Forwarded message ----------
> + > Date: Mon, 30 Oct 2006 13:44:28 +0530 (IST)
> + > From: Sabyasachi Sengupta <sabyasg@hp.com>
> + > To: Abhijit Toley <abhijit_toley@symantec.com>
> + > Cc: hpux-devtools@cxx.cup.hp.com, Amol Katkar
> + > <amol_katkar@symantec.com>,
> + > Prashant Gupta <Prashant_Gupta@symantec.com>
> + > Subject: Re: HPUX-DEVTOOLS: Important: Enquiry about "panicstr"
> + >
> + >
> + > Hi Abhijit,
> + >
> + > One point to remember:
> + >
> + > You can use arg1 = GSP_START_OF_SHUTDOWN coupled with arg2 =
> + > GSP_PANIC in gsp_register_callback when you want to register for
> + > getting notified upon a panic processing. For getting
> notified upon
> + > INIT/MCA, you could continue to use arg1 = GSP_CRASH and arg2 =
> + > appropriate mask (GSP_MCA | GSP_INIT).
> + >
> + > Both GSP_START_OF_SHUTDOWN and GSP_CRASH are called
> during a panic
> + > processing, but if you would want to get called at the
> start of the
> + > shutdown process upon a panic, GSP_START_OF_SHUTDOWN is what you
> + > would need. Also, if you want to get notified at the very
> start of
> + > panic processing, you could even register with GSP_EARLY_SHUTDOWN
> + > (with arg2 as GSP_PANIC). GSP_EARLY_SHUTDOWN is not possible for
> + > INIT's and MCA's.
> + >
> + > Thanks,
> + > Sabya
> + >
> + > --
> + >
> + > HPUX Kernel debugger tools project
> + > Ph: 91-80-22051892
> + > Mailstop: IE 2 Floor, Blue Wing-2, Cube 31
> + >
> + > On Mon, 30 Oct 2006, Sabyasachi Sengupta wrote:
> + >
> + > + Date: Mon, 30 Oct 2006 13:01:06 +0530 (IST)
> + > + From: Sabyasachi Sengupta <sabyasg@hp.com>
> + > + To: Abhijit Toley <abhijit_toley@symantec.com>
> + > + Cc: hpux-devtools@cxx.cup.hp.com,
> + > + Prashant Gupta <Prashant_Gupta@symantec.com>,
> + > + Amol Katkar <amol_katkar@symantec.com>
> + > + Subject: Re: HPUX-DEVTOOLS: Important: Enquiry about "panicstr"
> + > +
> + > +
> + > + Hi,
> + > +
> + > + I think you could look for the GSP shutdown primitives in
> + > + /usr/include/sys/gsp_callback_ksvc.h, refer to
> + > gsp_register_callback
> + > + and gsp_unregister_callback interfaces. Drivers could
> + > register to the
> + > + OS shutdown (crash) interface by specifying GSP_CRASH in
> + > param#1, an
> + > + appropriate mask (combination) in param#2, and
> GSP_KEEP_CALLBACK
> + > + or GSP_REMOVE_CALLBACK in param#3.
> + > +
> + > + This will ensure that HPUX will call your entrypoint
> upon a kernel
> + > + crash (panic, INIT/TOC, MCA/HPMC) instead of having to rely
> + > on panicstr. Note:
> + > + panicstr is set much later in the panic processing, so it is
> + > + better not to rely on it while scheduling further threads.
> + > +
> + > + Thanks,
> + > + Sabya
> + > +
> + > +
> + >
> + >
> + >
> + >
> + >
> + >
> + >
> + >
> + >
> + > _________________________________________________________________
> + > To leave this mailing list, send mail to majordomo@cxx.cup.hp.com
> + > with the message UNSUBSCRIBE hpux-devtools
> + > _________________________________________________________________
> + >
> +
>
>
> _________________________________________________________________
> To leave this mailing list, send mail to majordomo@cxx.cup.hp.com
> with the message UNSUBSCRIBE hpux-devtools
> _________________________________________________________________
>
_________________________________________________________________
To leave this mailing list, send mail to majordomo@cxx.cup.hp.com
with the message UNSUBSCRIBE hpux-devtools
_________________________________________________________________
[prev in list] [next in list] [prev in thread] [next in thread]
Configure |
About |
News |
Add a list |
Sponsored by KoreLogic