David S. Miller wrote:
From: Ben Greear [EMAIL PROTECTED]
Date: Mon, 29 Aug 2005 16:01:11 -0700
Latest netdevice ref-count debugging patch is up. The
patch is against 2.6.13:
http://www.candelatech.com/oss/rfcnt.patch
I reviewed this, and I think the approach can be refined and made more
From: Ben Greear [EMAIL PROTECTED]
Date: Thu, 01 Sep 2005 23:11:28 -0700
A quick optimization is to kmalloc chunks of 1000 or so structs at once
and then write a cheap foomalloc method to grab and release them. We
already take a lock (at least in my implementation), so this should be
small
David S. Miller wrote:
From: Ben Greear [EMAIL PROTECTED]
Date: Thu, 01 Sep 2005 23:11:28 -0700
A quick optimization is to kmalloc chunks of 1000 or so structs at once
and then write a cheap foomalloc method to grab and release them. We
already take a lock (at least in my implementation), so
David S. Miller wrote:
From: Ben Greear [EMAIL PROTECTED]
All the list stuff gets compiled out unless we're debugging.
So this netdev_pointer degenerates into a fancy pointer.
We can use a struct list_head to avoid the unlink complexity.
Ok, glad to see it can be compiled away as needed.
After some sleep, I have an idea:
Is there any reason we couldn't create a method:
void dev_hold_ref(struct dbg_dev_rfcnt_info* i, netdevice* dev, void* key);
This would be virtual identical to the __debug_dev_hold(netdevice* dev, void*
key)
that I currently have, but it would use the
From: Ben Greear [EMAIL PROTECTED]
Date: Fri, 02 Sep 2005 09:35:59 -0700
Is there any reason we couldn't create a method:
void dev_hold_ref(struct dbg_dev_rfcnt_info* i, netdevice* dev, void* key);
Please strongly consider my netdev_pointer idea,
and drop this dynamically allocated
From: Ben Greear [EMAIL PROTECTED]
Date: Mon, 29 Aug 2005 16:01:11 -0700
Latest netdevice ref-count debugging patch is up. The
patch is against 2.6.13:
http://www.candelatech.com/oss/rfcnt.patch
I reviewed this, and I think the approach can be refined and made more
robust.
The worst part
Latest netdevice ref-count debugging patch is up. The
patch is against 2.6.13:
http://www.candelatech.com/oss/rfcnt.patch
DaveM: Please let me know what you think about including
this for 2.6.14. Also, I was thinking we could probably always
compile for this feature but leave it disabled
From: Ben Greear [EMAIL PROTECTED]
Date: Mon, 29 Aug 2005 16:01:11 -0700
DaveM: Please let me know what you think about including
this for 2.6.14.
It'll take me a day or so to get back to this as I'm currently
getting net-2.6.14 merged into Linus's tree (some 240+ patches).
-
To unsubscribe
David S. Miller wrote:
From: Ben Greear [EMAIL PROTECTED]
Date: Sat, 27 Aug 2005 11:19:24 -0700
Although my implementation is not there yet, it would be nice if we
could use the address of the pointer than holds the dev reference
as the key. We should then be able to also catch double
On Sun, 2005-28-08 at 01:19 -0700, Ben Greear wrote:
Here's an updated patch (git working better now, thanks!):
http://www.candelatech.com/oss/rfcnt.patch
I started a list of constants #defined in netdevice.h. These
will be used for all stack variable references and other places
where I
From: Ben Greear [EMAIL PROTECTED]
Date: Fri, 26 Aug 2005 16:43:29 -0700
Btw, I think I found a leaked reference off in the ATM code...does this
look legit to you?
Definitely, that reference is leaked.
-
To unsubscribe from this list: send the line unsubscribe netdev in
the body of a message
David S. Miller wrote:
From: Ben Greear [EMAIL PROTECTED]
Date: Fri, 26 Aug 2005 16:43:29 -0700
Btw, I think I found a leaked reference off in the ATM code...does this
look legit to you?
Definitely, that reference is leaked.
Ok, I have the beginnings of my attempt at reference count
From: Ben Greear [EMAIL PROTECTED]
Date: Sat, 27 Aug 2005 01:23:25 -0700
I'm still learning git...I had to post filter that patch to get rid of
of the many diff --git lines that didn't seem to help much.
The timestamps of the files are stored in the git index, so if they
are different for
From: Ben Greear [EMAIL PROTECTED]
Date: Sat, 27 Aug 2005 01:23:25 -0700
The more interesting parts of the patch is below. The basic
idea is to pass a key in when you hold or put a device. This key
matches puts with holds. #defined macros grab the file name and
line number for printing out
David S. Miller wrote:
From: Ben Greear [EMAIL PROTECTED]
Date: Sat, 27 Aug 2005 01:23:25 -0700
The more interesting parts of the patch is below. The basic
idea is to pass a key in when you hold or put a device. This key
matches puts with holds. #defined macros grab the file name and
line
From: Ben Greear [EMAIL PROTECTED]
Date: Sat, 27 Aug 2005 11:19:24 -0700
Although my implementation is not there yet, it would be nice if we
could use the address of the pointer than holds the dev reference
as the key. We should then be able to also catch double dereferences
and places where
David S. Miller wrote:
From: Ben Greear [EMAIL PROTECTED]
Date: Sat, 27 Aug 2005 11:19:24 -0700
Although my implementation is not there yet, it would be nice if we
could use the address of the pointer than holds the dev reference
as the key. We should then be able to also catch double
In dev.c's process_backlog, there is code like this:
dev = skb-dev
netif_receive_skb(skb);
dev_put(dev);
Do you happen to know what piece of code takes this reference to
the device in the first place?
Also, why not make the
From: Ben Greear [EMAIL PROTECTED]
Date: Fri, 26 Aug 2005 16:34:49 -0700
And for the record, I grepped a great deal of dev_holds, but
associating a particular get with a put is making my head hurt :)
How about the one, and only one, in that file that adds the SKBs to
the input_pkt_queue list?
From: Ben Greear [EMAIL PROTECTED]
Date: Wed, 24 Aug 2005 10:14:22 -0700
I was thinking about somehow instrumenting dev_get and dev_put
(and __dev_get and __dev_put, I suppose) to keep a list of all
that have got and put, but that will be very invasive and time
consuming I fear...
That's how
21 matches
Mail list logo
