[9fans] tips debugging fs

[9fans] tips debugging fs

[Gabriel Diaz]

[-- Attachment #1: Type: text/plain, Size: 381 bytes --]

Hello,

following the example of /sys/lib/lib9p/ramfs.c,
i have done an fs test that dies unexpectedly with
no error message, just " i/o hungup channel " because
the proc died.

i know about memory leaks in there, and i'm trying to
found them with leak, but the proc dies just after running
leak -s.

how can i debug that? i shoud attach acid the first, and then
run leak?

thanks

[-- Attachment #2: Type: text/html, Size: 446 bytes --]

Re: [9fans] tips debugging fs

[Russ Cox]

If the process is not calling exits,
then the broken process should still
be around and debuggable.

ps | grep Broken

and then run acid on that pid.

Running leak on a program should
not change its behavior (it doesn't even
stop the program), so I don't think that's
causing the crash.

Russ

Re: [9fans] tips debugging fs

[andrey mirtchovski]

> ps | grep Broken

the command 'broke' is also convenient.

Re: [9fans] tips debugging fs

[Gabriel Diaz]

[-- Attachment #1: Type: text/plain, Size: 652 bytes --]

Hello

i have changed exits() to abort(), but when the error appears,
there is no broken proc,  so i suppose is srv.c
who is calling exits :-? but i didn't see any message
(perror should print the message isn't it ? ).

i will add some more prints to see where it die.

thanks

gabi


On 2/25/06, Russ Cox <rsc@swtch.com> wrote:
>
> If the process is not calling exits,
> then the broken process should still
> be around and debuggable.
>
> ps | grep Broken
>
> and then run acid on that pid.
>
> Running leak on a program should
> not change its behavior (it doesn't even
> stop the program), so I don't think that's
> causing the crash.
>
> Russ
>
>

[-- Attachment #2: Type: text/html, Size: 983 bytes --]

Re: [9fans] tips debugging fs

[Russ Cox]

> i will add some more prints to see where it die.

One tip is to run with -D (assuming that does chatty9p++)
and then watch the trace.  See which message is causing it
to die.

Another possibility is that you've removed the /srv file
and no one has it mounted, so it is simply exiting normally.

Russ

Re: [9fans] tips debugging fs

[Gabriel Diaz]

[-- Attachment #1: Type: text/plain, Size: 682 bytes --]

hello

i just test that and using test -s test -m /mnt/test i can
see a /srv/test, and when the clone fails /srv/test doesn't exist.

it crashed when i finish a read on a file and the last 9p messages are
Read
Tclunk
Rclunk

So I am removing a fid that i need and that is causing fs to exits normally
:-?

gabi


On 2/25/06, Russ Cox <rsc@swtch.com> wrote:
>
> > i will add some more prints to see where it die.
>
> One tip is to run with -D (assuming that does chatty9p++)
> and then watch the trace.  See which message is causing it
> to die.
>
> Another possibility is that you've removed the /srv file
> and no one has it mounted, so it is simply exiting normally.
>
> Russ
>
>

[-- Attachment #2: Type: text/html, Size: 1018 bytes --]