Udevd sequence number overlap issue!

Udevd sequence number overlap issue!

[Yin, Hu]

Hi, All,

I have a question, that is, why udevd didn't handle sequence number overlap issue---that means it's possible for udevsend to send two messages with the same sequence number to udevd, especially for test environment. I know udevd can execute successfully at this case. But should we explicitly handle this issue?

Should we take this issue into consideration? Look forward to your replying.

Best Regards,
 
Yin Hu ( Nick, a Intel Intern Student)



-------------------------------------------------------
This SF.Net email is sponsored by: IBM Linux Tutorials
Free Linux tutorial presented by Daniel Robbins, President and CEO of
GenToo technologies. Learn everything from fundamentals to system
administration.http://ads.osdn.com/?ad_id\x1470&alloc_id638&opÌk
_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

Re: Udevd sequence number overlap issue!

[Kay Sievers]

On Wed, Apr 07, 2004 at 09:40:55AM +0800, Yin, Hu wrote:
> Hi, All,
> 
> I have a question, that is, why udevd didn't handle sequence number
> overlap issue---that means it's possible for udevsend to send two
> messages with the same sequence number to udevd, especially for test
> environment. I know udevd can execute successfully at this case.
> But should we explicitly handle this issue?
> 
> Should we take this issue into consideration? Look forward to your replying.

If you change the sequences, like it happens while you run a test script
with hardcoded numbers, the first event will be delayed by 10 seconds.
You can prevent this by killing udevd before running the test.  It will
expect sequence number 0 as the first event.

You can also bypass the timeout by omitting SEQNUM for udevsend, it will
bypass the timeout handling and move straight to the exec queue.

If you want to test udev - bypass the timeout. If you want to test
udevsend/udevd just kill udevd or wait one timeout for reinitialization
of the exepected_seqnum.
What do you think is missing and why should we have a special test case?
In both cases you get exactly the expected behavior.

thanks,
Kay


-------------------------------------------------------
This SF.Net email is sponsored by: IBM Linux Tutorials
Free Linux tutorial presented by Daniel Robbins, President and CEO of
GenToo technologies. Learn everything from fundamentals to system
administration.http://ads.osdn.com/?ad_id\x1470&alloc_id638&op=click
_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

RE: Udevd sequence number overlap issue!

[Yin, Hu]

Kay,

First thank you very much for your reply. Maybe I didn't describe this
issue clearly. I understand what you have said. But I mean what it will
do when udevd receive two messages with the same sequence number at the
same time. Is it not necessary for udevd to check whether the incoming
sequence number has existed:-)

Thank you very much!

Yin Hu ( Nick )

-----Original Message-----
From: Kay Sievers [mailto:kay.sievers@vrfy.org] 
Sent: Wednesday, April 07, 2004 6:05 PM
To: Yin, Hu
Cc: linux-hotplug-devel@lists.sourceforge.net
Subject: Re: Udevd sequence number overlap issue!

On Wed, Apr 07, 2004 at 09:40:55AM +0800, Yin, Hu wrote:
> Hi, All,
> 
> I have a question, that is, why udevd didn't handle sequence number
> overlap issue---that means it's possible for udevsend to send two
> messages with the same sequence number to udevd, especially for test
> environment. I know udevd can execute successfully at this case.
> But should we explicitly handle this issue?
> 
> Should we take this issue into consideration? Look forward to your
replying.

If you change the sequences, like it happens while you run a test script
with hardcoded numbers, the first event will be delayed by 10 seconds.
You can prevent this by killing udevd before running the test.  It will
expect sequence number 0 as the first event.

You can also bypass the timeout by omitting SEQNUM for udevsend, it will
bypass the timeout handling and move straight to the exec queue.

If you want to test udev - bypass the timeout. If you want to test
udevsend/udevd just kill udevd or wait one timeout for reinitialization
of the exepected_seqnum.
What do you think is missing and why should we have a special test case?
In both cases you get exactly the expected behavior.

thanks,
Kay


-------------------------------------------------------
This SF.Net email is sponsored by: IBM Linux Tutorials
Free Linux tutorial presented by Daniel Robbins, President and CEO of
GenToo technologies. Learn everything from fundamentals to system
administration.http://ads.osdn.com/?ad_id\x1470&alloc_id638&opÌk
_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

Re: Udevd sequence number overlap issue!

[Kay Sievers]

On Thu, Apr 08, 2004 at 09:32:31AM +0800, Yin, Hu wrote:
> First thank you very much for your reply. Maybe I didn't describe this
> issue clearly. I understand what you have said. But I mean what it will
> do when udevd receive two messages with the same sequence number at the
> same time.

The kernel will _never_ send two messages with the same sequence. This
case is really pathological and affects only _your_ testing setup. It
has _nothing_ to do with the real world.

> Is it not necessary for udevd to check whether the incoming
> sequence number has existed:-)

Is this a question? Sure it's neccessary. How will you reorder incoming
sequences without remembering the number of the last handled sequence?
Maybe I don't get your point?

Kay


-------------------------------------------------------
This SF.Net email is sponsored by: IBM Linux Tutorials
Free Linux tutorial presented by Daniel Robbins, President and CEO of
GenToo technologies. Learn everything from fundamentals to system
administration.http://ads.osdn.com/?ad_id\x1470&alloc_id638&op=click
_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel